CN111695922B - Potential user determination method and device, storage medium and electronic device - Google Patents

Disclosure of Invention

The present disclosure aims to provide a potential user determination method, a potential user determination device, a storage medium and an electronic apparatus, thereby overcoming, at least to some extent, the problem of low accuracy of potential user identification due to the related art.

According to one aspect of the disclosure, a potential user determining method is provided, which comprises the steps of determining a plurality of item information according to item interaction behaviors of a target user, respectively converting each item information into an item vector, constructing a first item sequence composed of the plurality of item vectors, inputting the first item sequence into a trained intention item prediction model to determine a second item sequence, determining a target item based on the second item sequence and determining item acquisition characteristics of the target user in combination with the information of the target item, inputting the item acquisition characteristics into a trained probability model to determine target item acquisition probability of the target user, and determining whether the target user is a potential user for acquiring the target item according to the target item acquisition probability.

In an exemplary embodiment of the disclosure, converting each item of information into an item of vector includes performing word segmentation on the item of information to obtain each word segmentation result, inputting each word segmentation result into a trained first conversion model to determine a word segmentation vector corresponding to each word segmentation result, and integrating each word segmentation vector to determine an item of vector corresponding to the item of information.

In one exemplary embodiment of the present disclosure, converting each item information into an item vector includes determining a number of browses of items corresponding to each item information to determine a browsing sequence of each item, and inputting the browsing sequence into a trained second conversion model to determine an item vector corresponding to each item information.

In one exemplary embodiment of the disclosure, the potential user determination method comprises the steps of determining a plurality of first users, determining a sample input sequence according to a plurality of item information determined by item interaction behaviors of the first users, determining a sample output sequence according to intention item information of the first users, taking the sample input sequence as input of an intention item prediction model, and taking the sample output sequence as output of the intention item prediction model, so as to train the intention item prediction model.

In one exemplary embodiment of the present disclosure, determining a target item based on the second item sequence includes taking an item contained in the second item sequence as a target item.

In one exemplary embodiment of the present disclosure, determining a target item based on the second item sequence includes determining an item contained in the second item sequence as a first candidate item, determining an item similar to the first candidate item as a second candidate item, and determining a set of the first candidate item and the second candidate item as a target item.

In one exemplary embodiment of the disclosure, determining that an item similar to the first candidate item is a second candidate item includes calculating a similarity of the first candidate item to other items in the set of items, and determining that an item in the set of items having a similarity greater than a preset similarity threshold is a second candidate item.

In one exemplary embodiment of the present disclosure, determining the item acquisition characteristics of the target user in conjunction with the information of the target item includes determining a first characteristic of the target user based on item interaction behavior of the target user over a first preset period of time, determining a second characteristic of the target user based on a user representation of the target user, extracting base information from the information of the target item and determining the first characteristic of the target item based on the base information, extracting dynamic information from the information of the target item and determining the second characteristic of the target item based on the dynamic information, and combining the first and second characteristics of the target user and the first and second characteristics of the target item to determine the item acquisition characteristics of the target user.

In one exemplary embodiment of the disclosure, the potential user determination method further includes determining a plurality of second users, determining item acquisition characteristics of each of the second users, determining user information of presence intention behaviors within a second preset time period, determining training samples of a probability model based on the item acquisition characteristics of each of the second users and the user information of presence intention behaviors within the second preset time period, and training the probability model by using the training samples of the probability model.

In one exemplary embodiment of the present disclosure, determining the item acquisition characteristics of each of the second users includes determining a first characteristic of the second user according to the item interaction behavior of the second user within a third preset time period, determining a second characteristic of the second user based on a user image of the second user, extracting basic information from information of a predetermined item and determining the first characteristic of the predetermined item according to the basic information, extracting dynamic information from the information of the predetermined item and determining a second characteristic of the predetermined item according to the dynamic information, and combining the first and second characteristics of the second user and the first and second characteristics of the predetermined item to determine the item acquisition characteristics of the second user.

In one exemplary embodiment of the present disclosure, determining whether the target user is a potential user to acquire the target item based on the target item acquisition probability includes determining whether the target item acquisition probability is greater than a preset probability threshold, and if the target item acquisition probability is greater than the preset probability threshold, determining the target user as a potential user to acquire the target item.

In one exemplary embodiment of the present disclosure, after determining that the target user is a potential user for obtaining the target item, the user determining method further includes pushing first preference information to the target user if the target user has intention achievement behavior for the target item within a fourth preset time period, and pushing second preference information to the target user if the target user has no intention achievement behavior for the target item within the fourth preset time period.

According to one aspect of the present disclosure, a potential user determination device is provided that includes an item sequence construction module, an intent item prediction module, an acquisition feature determination module, and a potential user determination module.

The object sequence construction module is used for determining a plurality of object information according to object interaction behaviors of a target user, converting the object information into object vectors respectively, constructing a first object sequence composed of the object vectors, the object prediction module is used for inputting the first object sequence into a trained object prediction model to determine a second object sequence, the acquisition characteristic determination module is used for determining a target object based on the second object sequence and determining object acquisition characteristics of the target user in combination with the information of the target object, and the potential user determination module is used for inputting the object acquisition characteristics into a trained probability model to determine target object acquisition probability of the target user and determining whether the target user is a potential user for acquiring the target object according to the target object acquisition probability.

In one exemplary embodiment of the present disclosure, the item sequence construction module includes a first conversion unit.

The first conversion unit is configured to execute word segmentation processing on the article information to obtain word segmentation results, input the word segmentation results into a trained first conversion model to determine word segmentation vectors corresponding to the word segmentation results, and integrate the word segmentation vectors to determine article vectors corresponding to the article information.

In one exemplary embodiment of the present disclosure, the item sequence construction module includes a second conversion unit.

Specifically, the second conversion unit is configured to determine the browsing times of the articles corresponding to the article information so as to determine the browsing sequence of the articles, and input the browsing sequence into a trained second conversion model so as to determine the article vector corresponding to the article information.

In one exemplary embodiment of the present disclosure, the potential user determination device further comprises a first model training module.

The first model training module is configured to determine a plurality of first users, determine a sample input sequence according to a plurality of item information determined by item interaction behaviors of the first users, determine a sample output sequence according to the intended item information of the first users, take the sample input sequence as input of an intended item prediction model, and take the sample output sequence as output of the intended item prediction model so as to train the intended item prediction model.

In one exemplary embodiment of the present disclosure, the acquisition characteristic determination module includes a first target item determination unit.

Specifically, the first target article determining unit is configured to take an article contained in the second article sequence as a target article.

In one exemplary embodiment of the present disclosure, the acquisition characteristic determination module includes a second target item determination unit.

Specifically, the second target item determining unit is configured to determine items contained in the second item sequence as first candidate items, determine items similar to the first candidate items as second candidate items, and determine a set of the first candidate items and the second candidate items as target items.

In one exemplary embodiment of the present disclosure, the second target item determination unit includes a similarity calculation subunit and an item determination subunit.

The article determining subunit is used for taking the article with the similarity in the article set being greater than a preset similarity threshold value as a second candidate article.

In one exemplary embodiment of the present disclosure, the acquisition characteristic determination module includes a first acquisition characteristic determination unit.

Specifically, the first acquisition characteristic determining unit is configured to determine a first characteristic of the target user according to article interaction behaviors of the target user in a first preset time period, determine a second characteristic of the target user based on a user portrait of the target user, extract basic information from information of the target article and determine the first characteristic of the target article according to the basic information, extract dynamic information from the information of the target article and determine the second characteristic of the target article according to the dynamic information, and combine the first characteristic and the second characteristic of the target user and the first characteristic and the second characteristic of the target article to determine the article acquisition characteristic of the target user.

In one exemplary embodiment of the present disclosure, the potential user determination device further comprises a second model training module.

Specifically, the second model training module is configured to perform determining a plurality of second users, determining object acquisition characteristics of each of the second users, determining user information of presence intention behaviors within a second preset time period, determining training samples of a probability model based on the object acquisition characteristics of each of the second users and the user information of presence intention behaviors within the second preset time period, and training the probability model by using the training samples of the probability model.

In an exemplary embodiment of the present disclosure, the second model training module includes a second acquisition feature determination unit.

Specifically, the second acquisition characteristic determining unit is configured to determine a first characteristic of the second user according to the article interaction behavior of the second user in a third preset time period, determine a second characteristic of the second user based on the user image of the second user, extract basic information from information of a preset article and determine the first characteristic of the preset article according to the basic information, extract dynamic information from the information of the preset article and determine the second characteristic of the preset article according to the dynamic information, and combine the first characteristic and the second characteristic of the second user and the first characteristic and the second characteristic of the preset article to determine the article acquisition characteristic of the second user.

In one exemplary embodiment of the present disclosure, the potential user determination module includes a potential user determination unit.

Specifically, the potential user determining unit is configured to determine whether the target object acquisition probability is greater than a preset probability threshold, and determine the target user as a potential user for acquiring the target object if the target object acquisition probability is greater than the preset probability threshold.

In one exemplary embodiment of the present disclosure, the potential user determination device further comprises a coupon information push module.

Specifically, the offer information pushing module is configured to push first offer information to the target user if the target user has intention achievement behavior for the target object within a fourth preset time period, and push second offer information to the target user if the target user does not have intention achievement behavior for the target object within the fourth preset time period.

According to one aspect of the present disclosure, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, implements the potential user determination method of any of the embodiments described above.

According to one aspect of the disclosure, an electronic device is provided, comprising a processor, and a memory for storing executable instructions of the processor, wherein the processor is configured to perform the potential user determination method of any of the embodiments described above via execution of the executable instructions.

In some embodiments of the present disclosure, a first item sequence determined based on item interaction of a target user is input into an intended item prediction model to obtain a second item sequence representing a possible interest of the user, a target item is determined based on the second item sequence, item acquisition characteristics of the target user are determined in combination with information of the target item, the item acquisition characteristics are input into a probability model to obtain a target item acquisition probability of the target user, and whether the target user is a potential user for acquiring the target item is determined according to the target item acquisition probability. On one hand, the method and the device for determining the electronic commerce platform can accurately determine potential users by combining the article interaction behaviors of the users, further help to guide the behaviors of the potential users, save the time of searching articles for the users, on the other hand, the method and the device are beneficial to analysis, calculation and processing by using a model by converting article information into a vector form, on the other hand, the method and the device for determining the object by using a second article sequence which possibly interests the object users, then determine the object acquisition characteristics according to the object, and input the object acquisition characteristics into a probability model to determine whether the object users are potential users or not, therefore, the problem of identifying the potential users can be solved better, the potential user positioning result with better flexibility can be provided by using a probability mode, and the operation effect of the electronic commerce platform is improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein, but rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the present disclosure. One skilled in the relevant art will recognize, however, that the aspects of the disclosure may be practiced without one or more of the specific details, or with other methods, components, devices, steps, etc. In other instances, well-known technical solutions have not been shown or described in detail to avoid obscuring aspects of the present disclosure.

Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor devices and/or microcontroller devices.

The flow diagrams depicted in the figures are exemplary only and not necessarily all steps are included. For example, some steps may be decomposed, and some steps may be combined or partially combined, so that the order of actual execution may be changed according to actual situations. In addition, the terms "first," "second," "third," "fourth," etc. are used herein for distinguishing purposes only and should not be taken as a limitation of the present disclosure.

The short message and the e-mail are good ways for promoting products of the e-commerce platform, the products can correspond to various articles, and the articles can be promoted to users in a paid or gratuitous way. The short message and the E-mail have the characteristics of low cost and high touch rate. However, in recent years, because short messages and emails are abused, popularization of the electronic commerce platform is blocked, and an effect of guiding user behaviors cannot be well achieved.

At present, the more common method for determining potential users can comprise the steps of mining similar users through user portraits and relationship chains based on a Lookalike model of seed population expansion, specifically based on pre-defined seed populations, determining a user method based on user labels, specifically using search behavior information of the user to determine the user labels and determining the potential users according to the user labels, generating keyword feature vectors based on user search and access results, extracting feature vectors of target objects according to texts, and mining the potential users by using a neighbor search method. In addition, there are some methods based on human defined metrics and calculation rules.

However, the Lookalike model has wider directional crowd, lacks the behavior characteristics of users, is difficult for an electronic commerce platform to actively touch potential users who may acquire products, has lower conversion rate, is difficult to process the problem related to sequencing besides lacking the behavior characteristics of the users, and ignores part of operation behaviors of the users by an unsupervised method based on text analysis, and has the defects of great consumption of human resources, poor universality of the method and manual definition indexes and calculation rules. Thus, it can be seen that the existing processing method has the defect of low accuracy of potential user identification.

In view of this, the exemplary embodiments of the present disclosure provide a new method and apparatus for determining a potential user, which can improve accuracy of identifying the potential user, and thus help to guide user behavior.

It should be noted that the potential user determination method of the exemplary embodiments of the present disclosure may be implemented by a server, that is, the server may implement the steps of the potential user determination method of the present disclosure. For example, the server may be a server used by the e-commerce platform, and in addition, the server may be a third party server serving the e-commerce platform, which is not particularly limited in this exemplary embodiment. In this case, the potential user determination means described below may be configured within the server.

Fig. 1 schematically illustrates a flow chart of a potential user determination method of an exemplary embodiment of the present disclosure. Referring to fig. 1, the potential user determination method may include the steps of:

S12, determining a plurality of item information according to item interaction behaviors of a target user, respectively converting each item information into an item vector, and constructing a first item sequence consisting of the plurality of item vectors.

In an exemplary embodiment of the present disclosure, the target user is a user to be determined whether it is a potential user, for example, a developer may construct a user whitelist in advance, and the target user may be any user in the user whitelist. Additionally, the user's item interaction behavior may include, but is not limited to, user browsing behavior of the item, behavior of the item being added to the acquisition bar, behavior of the user focusing on the item, user pre-acquisition behavior of the item, user collection behavior of the item, and the like. Wherein the pre-acquisition behavior may also be referred to as intent behavior, characterizing the user's interest in acquiring the item.

According to some embodiments of the present disclosure, the server may determine the plurality of item information from one of the item interactions listed above.

For example, a plurality of item information may be determined according to browsing behavior. Specifically, first, the server may extract browsing behavior data from a flow table in a data warehouse storing user data, and on one hand, the server may extract browsing behavior data over a period of time (e.g., 30 days), and on the other hand, the browsing behavior data may be characterized as a triplet, for example, < user identification, browsing time, item ID >. Next, the server may determine a plurality of item information involved from these browsing behavior data, wherein the item information may include an item name, description information of the item, an item ID, item provider information, and the like.

It will be readily appreciated that similarly, the server may also determine a plurality of item information based on the item's behavior data being added to the acquisition column, the user's behavior data concerning the item, the user's pre-acquisition behavior data for the item, or the user's collection behavior data for the item.

In addition, the server may determine a plurality of item information in combination with a plurality of the above-described behavior data, for example, the server may determine a plurality of item information in combination with user's attention behavior data on items and user's collection behavior data on items, which is not particularly limited in this exemplary embodiment.

After determining the plurality of item information, the server may convert each item information into an item vector, respectively, for later analysis processing.

According to some embodiments of the present disclosure, first, the server may perform word segmentation processing on each item information to obtain word segmentation results. Specifically, the existing natural language processing method may be used to segment the article information, the specific implementation manner of the segmentation is not limited in this disclosure, and then the server may input each obtained segmentation result into a trained first conversion model to determine a segmentation vector of each segmentation result, where the dimension of the segmentation vector may be, for example, 64. The first transformation model may be a Word2vev model, however, the first transformation model may also be GloVe models, ELMo models, BERT models, etc., and the disclosure does not limit the type of the first transformation model, the selection of the training samples, and the training process, and then, the server may integrate the analysis vectors output via the first transformation model to determine the article vectors corresponding to the article information.

According to other embodiments of the present disclosure, first, the server may determine the browsing times of the items corresponding to each item information, and determine the browsing sequence of each item based on the browsing times, for example, the browsing sequence may be a sequence in which the items are arranged from high to low based on the browsing times, or may be a sequence in which the items are arranged from low to high based on the browsing times, which is not particularly limited in this exemplary embodiment, and then, the server may input the determined browsing sequence into a trained second conversion model, where the output of the second conversion model is an item vector corresponding to each item information. The second transformation model may be the same type as the first transformation model and be a Word2vev model, however, the second transformation model may also be different from the first transformation model and may be a Word2vev model, gloVe model, ELMo model, BERT model, or the like.

In addition, for an item information, the two embodiments can be combined to determine an item vector. For example, when the first embodiment determines that the dimension of the item vector is 64 and the second embodiment determines that the dimension of the item vector is 64, the two may be combined into a 128-dimensional vector to characterize the item information.

After converting each item information into an item vector, the server may include a first sequence of items consisting of a plurality of item vectors. For example, in the case where the item interaction behavior is a browsing behavior, the first item sequence may be a sequence in which items are arranged from high to low in the number of browsing, or may be a sequence in which items are arranged from low to high in the number of browsing.

It should be appreciated that the scheme described above is a scheme for constructing a first sequence of items after converting each item information into an item vector. However, in other embodiments of the present disclosure, a sequence of item information may be constructed first, and then vector conversion may be performed on all item information based on the sequence, so as to determine the first item sequence.

Furthermore, the first conversion model and/or the second conversion model described above may be configured within the server. However, it is easily understood that the first conversion model and/or the second conversion model may also be configured in another server, which is not particularly limited in the present exemplary embodiment.

S14, inputting the first article sequence into a trained intention article prediction model to determine a second article sequence.

In an exemplary embodiment of the present disclosure, the intent-item prediction model may be a sequence-to-sequence based neural network model (seq 2seq neural network model), the seq2seq neural network model being an end-to-end (end-to-end) model that enables a prediction process to be performed on a sequence. The exemplary embodiments of the present disclosure employ a seq2seq neural network model to implement the process of intent item prediction, however, it should be understood that the seq2seq neural network model is not limited to a form, e.g., a seq2seq neural network model with an attention mechanism may also be employed. In addition, in other embodiments of the present disclosure, other neural network models may be used to process the first article sequence, which is not particularly limited in this exemplary embodiment. Similar to the first and second transformation models described above, the intent item prediction model may be configured within a server that performs the exemplary methods of the present disclosure, however, may also be configured within other servers than the server that performs the exemplary methods of the present disclosure.

Exemplary embodiments of the present disclosure may also include a process of training the intent-item predictive model.

First, the server may determine a plurality of first users, specifically, may determine that a user having pre-acquisition activity (intention to reach activity) for a period of time is the first user based on the flow table of the data warehouse storing the user data, however, may also determine the first user based on the user whitelist described in step S12, for example, a number of users may be selected randomly from the user whitelist as the first user, in which case the number of users in the user whitelist may be determined in combination with the computer processing capability.

Next, the server may determine a plurality of item information according to the item interaction behavior of each first user, and it should be understood that, for each first user of the plurality of first users, the plurality of item information may be determined according to the item interaction behavior of the first user, and a manner of determining a plurality of items corresponding to the first user is the same as a manner of determining a plurality of item information corresponding to the target user in step S12, which is not described herein again. After determining a plurality of item information corresponding to the first user, a sample input sequence may be determined according to the plurality of item information, and similarly, the plurality of item information may be converted into item vectors, respectively, to construct a sequence composed of the plurality of item vectors as the sample input sequence. It should be noted that for each first user there may be a corresponding sample input sequence.

In addition, the server may obtain the intent item information for each first user from the data warehouse and determine a sample output sequence based on the intent item information. Wherein the individual intent items may be ordered based on the number of intent items to generate a sample output sequence. Also, for each first user, there may be a corresponding sample output sequence. It should be appreciated that the sample output sequence is also represented in terms of an item vector.

It should be noted that the server in determining the sample input sequence and the sample output sequence is item information extracted over a time period, which may be, for example, one month, one quarter, one half year, one year, etc. It is readily understood that the larger the time period span, the larger the sample data, and the better the model predictive effect.

The server may then take the sample input sequence as input to the intent item prediction model and the sample output sequence as output to the intent item prediction model to train the intent item prediction model.

Further, after training the intent-item prediction model, the methods of the exemplary embodiments of the present disclosure may further include obtaining a sample for verifying the intent-item prediction model in order to verify the trained intent-item prediction model. The process of obtaining a sample of the verification model and the process of verifying the model are similar to the above training process, and will not be described in detail herein.

The foregoing describes a scenario for training a predictive model of an item of interest by a server of an exemplary method of the present disclosure. However, according to other embodiments of the present disclosure, the process of training the intent-item prediction model may be performed by a server other than the server performing the exemplary methods of the present disclosure, and the training methods are similar and will not be described in detail herein.

After determining the trained intent item prediction model, the server may input the first item sequence constructed in step S12 into the trained intent item prediction model to determine the second item sequence. It is readily appreciated that the second item sequence characterizes an item sequence that may be of interest to the target user as determined based on the item interaction behavior of the target user.

S16, determining a target object based on the second object sequence, and determining object acquisition characteristics of the target user by combining information of the target object.

According to some embodiments of the present disclosure, the server may determine the items contained in the sequence from the second sequence of items and take these items as target items.

According to further embodiments of the present disclosure, a server may first determine items included in a sequence from a second sequence of items and treat the items as first candidate items, and next, the server may determine items similar to the first candidate item as second candidate items and determine a set of the first candidate item and the second candidate item as target items.

Specifically, the process of determining the target item may include:

first, the server may calculate a similarity of the first candidate item to each other item in the set of items.

The article set according to the exemplary embodiment of the present disclosure may be an article set made up of all articles of the electronic commerce platform, or may be an article set made up of all articles under a certain category.

The present disclosure may calculate similarity by cosine similarity principle. In particular, in view of the fact that the second article sequence is made up of article vectors of a plurality of articles, for example, one of the article vectors is noted asEach article in the article set is expressed as a vector form, and one article is selected and marked asFrom this, the similarity sim (X, Y) of two items can be calculated using the following formula:

thus, the similarity between the first candidate item and each other item in the item set can be calculated.

In addition to cosine similarity, further embodiments of the present disclosure may also calculate similarity between articles using, for example, the principles of mahalanobis distance, manhattan distance, euclidean distance, etc., which are not particularly limited in this exemplary embodiment.

Then, the server may use the item whose determined similarity is greater than a preset similarity threshold as the second candidate item, and for example, the preset similarity threshold may be set to 0.8. It will be readily appreciated that each first candidate item may correspond to a plurality of second candidate items.

The server may then determine a set of the first candidate item and the second candidate item as the target item.

After determining the target item, the server may determine an item acquisition characteristic of the target user in combination with information of the target item.

In one aspect, the server may determine the first characteristic of the target user based on the object interaction behavior of the target user over a first predetermined period of time (e.g., one day, one week, one month, one quarter, etc.). Taking the article interaction behavior as an example, the server can determine the article browsing sequence of the target user, count the browsing times of the target user for each article and the total browsing times of the target user, and take the browsing times of the target user for each article and the total browsing times of the target user as the first characteristics of the target user.

However, it should be understood that the item interaction behavior may be a combination of one or more of a browsing behavior of the item by the target user, a behavior of the item being added to the acquisition bar, a behavior of the target user focusing on the item, a pre-acquisition behavior of the item by the target user, a collection behavior of the item by the target user. Similarly, the data determined accordingly is the first feature, and for example, the number of times the target user joins the acquisition column for each item and the total number of times the target user joins the acquisition column may be taken as the first feature of the target user.

On the other hand, the server may determine the second characteristic of the target user based on the user representation of the target user. Specifically, the user portraits of the individual users may be pre-constructed and stored in a user portrayal table of the data warehouse. In this case, the server may first determine enumeration-type features, such as user account number, gender, age, membership grade, region, item acquisition capability grade, etc., from the user representation of the target user, then may uniorily (OneHot) encode these enumeration-type features to generate a plurality of 0/1 features, in addition, the server may extract numeric-type features, such as user promotion sensitivity scores, user liveness scores, etc., from the user representation of the target user, and then may splice the numeric-type features with the uniorily encoded enumeration-type features to obtain a second feature of the target user.

In yet another aspect, the server may extract base information from the information of the target item and determine the first characteristic of the target item based on the base information. Specifically, the server may first extract enumeration type features of the target item, such as item category, brand, color, intended user gender, etc., from a pre-stored item feature table, then may unithermally encode these enumeration type features to generate a plurality of 0/1 features, in addition, the server may extract numerical type features of the target item, such as size, weight, etc., from the item feature table, and then may splice the numerical type features of the target item with the unithermally encoded enumeration type features to obtain a first feature of the target item.

In yet another aspect, the server may extract dynamic information from the information of the target item and determine the second characteristic of the target item based on the dynamic information. Specifically, the server may extract, from a flow table of the data warehouse, browsed data, data added to an acquisition column, data concerned, collected data, data with intention behavior, and the like of the target item within a first preset period, and then determine corresponding values, for example, the browsed times, the times accessed by independent visitors (Unique Visitor, UV), the times added to the acquisition column, the times concerned, the times collected, the times with intention behavior, and the like, according to the data, and in addition, the server may obtain the amount of the target item emitted within the first preset period, and similarly, the server may determine the category or brand corresponding to the target item, and count the total browsed times, the times accessed by independent visitors, the times added to the acquisition column, the times concerned, the times collected, the times with intention behavior, the amount of emission, and the like. And integrates the information to determine a second characteristic of the target item.

Thus, the first and second characteristics of the target user and the first and second characteristics of the target item described above may be combined to determine the item acquisition characteristics of the target user.

S18, inputting the object acquisition characteristics into a trained probability model to determine the object acquisition probability of the object user, and determining whether the object user is a potential user for acquiring the object according to the object acquisition probability.

In an exemplary embodiment of the present disclosure, the probabilistic model may be a gradient-lifting decision tree model (Gradient Boosting Decision Tree, GBDT), and GBDT is an iterative decision tree algorithm that may be made up of multiple decision trees, with the conclusions of all trees being added to ultimately yield the predicted result. However, it should be understood that other probabilistic models may also be employed to implement the probabilistic models described in the present disclosure, e.g., random forests, XGBoost, etc., and the present disclosure is not limited in particular to the probabilistic models. Similar to the models described above, the probabilistic model of the exemplary embodiment of the present disclosure may be configured in a server that performs the exemplary method of the present disclosure, however, may be configured in other servers than the server that performs the exemplary method of the present disclosure.

Exemplary embodiments of the present disclosure may also include a process of training the probabilistic model.

First, the server may determine a plurality of second users. Similar to the determination of the plurality of first users in step S14, in particular, the server may determine the second user based on the behavior of the flow table of the data warehouse storing the user data over a period of time, however, the second user may also be determined based on the user whitelist described in step S12. It will be readily appreciated that the first user described above may be included in the plurality of second users.

Next, the server may determine the item acquisition characteristics of each second user separately, e.g., the server may determine the item acquisition characteristics of each second user daily.

The method comprises the steps of determining a first characteristic of a second user according to article interaction behaviors of the second user in a third preset time period, determining a second characteristic of the second user based on a user image of the second user, extracting basic information from information of a preset article, determining the first characteristic of the preset article according to the basic information, extracting dynamic information from the information of the preset article, determining the second characteristic of the preset article according to the dynamic information, and combining the first characteristic and the second characteristic of the second user and the first characteristic and the second characteristic of the preset article to determine article acquisition characteristics of the second user. The third preset time period may be one day, one week, one month, one quarter, or the like, and may be the same as the first preset time period, and in addition, the predetermined item may be a randomly determined item. The specific process is similar to the above-mentioned item acquisition feature of the target user determined in step S16, and will not be described here again.

In addition, the server may determine the user information of the presence intent behavior within a second preset time period, which may be a period of time (e.g., 15 days) after determining the second user item acquisition feature, however, the present disclosure does not specifically limit the second preset time period.

Then, training samples of the probability model may be determined based on the item acquisition characteristics of each second user and the user information of the presence intention behavior within the second preset period of time, and the probability model may be trained using the training samples.

For the process of determining the training samples of the probabilistic model, for example, the server may determine the item acquisition characteristics of each second user on 1 month and 1 day, and in the case that the second preset period is 15 days, determine the user information of the presence intention behavior on 1 month and 1 day to 1 month and 15 days. If the article acquisition characteristics and the user information of the existence intention behaviors are stored in a table manner, the article acquisition characteristics table and the user information table of the existence intention behaviors can be associated in a left connection manner, the data matched with the user information table of the existence intention behaviors in the article acquisition characteristics table is marked as a positive example, and the data not matched with the user information table of the existence intention behaviors is marked as a negative example. Next, the same number of positive examples and negative examples may be determined using a negative sampling method to obtain training samples for the 1 month 1 day probability model. Repeating the step of determining training samples for 1 month and 1 day for 1 month and 2 days to 1 month and 30 days to obtain training samples of the probability model.

And training the probability model by using the determined training sample. Additionally, after training the probabilistic model, the method of an exemplary embodiment of the present disclosure may further include obtaining a sample for validating the probabilistic model to validate the trained probabilistic model. The specific process of obtaining the sample of the verification probability model and the process of verifying the model are similar to the above-mentioned training process, and will not be described here again.

The foregoing describes a scenario for training a probabilistic model by a server of an exemplary method of the present disclosure. However, according to other embodiments of the present disclosure, the process of training the probabilistic model may be performed by other servers than the server performing the exemplary methods of the present disclosure, and the training methods are similar and will not be described in detail herein.

After determining the trained probability model, the server may input the item acquisition characteristics of the target user determined in step S16 into the trained probability model to determine a target item acquisition probability of the target user.

In addition, after determining the target item acquisition probability of the target user, the server may determine whether the target user is a potential user for acquiring the target item according to the target item acquisition probability. Specifically, the server may determine whether the target object acquisition probability is greater than a preset probability threshold, for example, the preset probability threshold may be set to 0.5. If the target item acquisition probability is greater than the preset probability threshold, the target user is determined to be a potential user for acquiring the target item.

For the determined potential user, the exemplary embodiments of this disclosure also provide a user behavior guidance policy. Specifically, if the target user has the intention achievement behavior for the target object within a fourth preset time period (for example, within half a year from the current time), the first preference information is pushed to the target user, and if the target user does not have the intention achievement behavior for the target object within the fourth preset time period, the second preference information is pushed to the target user.

And judging whether the target user has the intention achievement behaviors aiming at the target object in a fourth preset time period, classifying the target user, and carrying out different user behavior guiding strategies. For example, if the target user has intent to reach an action for the target item for a fourth preset time period, the target user may be indicated as an old user, in which case first offer information may be pushed for the target user based on the policy for the old user, e.g., credit doubling, coupon issue, etc., and if the target user does not have intent to reach an action for the target item for the fourth preset time period, the target user may be indicated as a new user, and second offer information may be pushed for the target user based on the policy for the new user, e.g., coupon issue. It is readily understood that the denomination of the coupons issued for the old and new users may be different.

It should be noted that although the steps of the methods in the present disclosure are depicted in the accompanying drawings in a particular order, this does not require or imply that the steps must be performed in that particular order, or that all illustrated steps be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform, etc.

Further, a potential user determining apparatus is provided in the present exemplary embodiment.

Fig. 2 schematically illustrates a block diagram of a potential user determination device of an exemplary embodiment of the present disclosure. Referring to fig. 2, a potential user determination device 2 according to an exemplary embodiment of the present disclosure may include an item sequence construction module 21, an intent item prediction module 23, an acquisition feature determination module 25, and a potential user determination module 27.

Specifically, the article sequence construction module 21 may be configured to determine a plurality of article information according to article interaction behaviors of a target user, respectively convert each article information into an article vector, and construct a first article sequence composed of a plurality of article vectors, the intent article prediction module 23 may be configured to input the first article sequence into a trained intent article prediction model to determine a second article sequence, the acquisition feature determination module 25 may be configured to determine a target article based on the second article sequence and determine an article acquisition feature of the target user in combination with the information of the target article, and the potential user determination module 27 may be configured to input the article acquisition feature into a trained probability model to determine a target article acquisition probability of the target user, and determine whether the target user is a potential user who acquires the target article according to the target article acquisition probability.

According to the potential user determining device based on the exemplary embodiment of the disclosure, on one hand, the potential user can be accurately determined by combining with the article interaction behavior of the user, the article searching time is saved for the user, on the other hand, analysis and processing by using a model are facilitated by converting article information into a vector form, on the other hand, a target article is determined by using a second article sequence which is possibly interested by the target user, then the article acquisition characteristics are determined according to the target article, and the article acquisition characteristics are input into a probability model to determine whether the target user is the potential user or not, so that the identification problem of the potential user can be better solved, the potential user positioning result with better flexibility can be provided by using a probability mode, and the operation effect of the electronic commerce platform is improved.

According to an exemplary embodiment of the present disclosure, referring to fig. 3, the item sequence construction module 21 may include a first conversion unit 301.

Specifically, the first conversion unit 301 may be configured to perform word segmentation on the article information to obtain word segmentation results, input each word segmentation result into a trained first conversion model to determine a word segmentation vector corresponding to each word segmentation result, and integrate each word segmentation vector to determine an article vector corresponding to the article information.

According to an exemplary embodiment of the present disclosure, referring to fig. 4, the item sequence construction module 21 may include a second conversion unit 401.

Specifically, the second conversion unit 401 may be configured to determine the browsing times of the items corresponding to the item information to determine the browsing sequence of the items, and input the browsing sequence into a trained second conversion model to determine the item vector corresponding to the item information.

According to an exemplary embodiment of the present disclosure, referring to fig. 5, the potential user determination device 5 may include a first model training module 51 in addition to the item sequence construction module 21, the intent item prediction module 23, the acquisition characteristic determination module 25, and the potential user determination module 27, as compared to the potential user determination device 2.

Specifically, the first model training module 51 may be configured to perform determining a plurality of first users, determining a sample input sequence according to a plurality of item information determined by item interaction behaviors of each of the first users, and determining a sample output sequence according to intended item information of each of the first users, taking the sample input sequence as an input of an intended item prediction model, and taking the sample output sequence as an output of the intended item prediction model, so as to train the intended item prediction model.

According to an exemplary embodiment of the present disclosure, referring to fig. 6, the acquisition characteristic determining module 25 may include a first target item determining unit 601.

Specifically, the first target item determining unit 601 may be configured to take an item included in the second item sequence as a target item.

According to an exemplary embodiment of the present disclosure, referring to fig. 7, the acquisition characteristic determining module 25 may include a second target item determining unit 701.

Specifically, the second target item determining unit 701 may be configured to perform determining items included in the second item sequence as first candidate items, determining items similar to the first candidate items as second candidate items, and determining a set of the first candidate items and the second candidate items as target items.

According to an exemplary embodiment of the present disclosure, referring to fig. 8, the second target item determination unit 701 may include a similarity calculation subunit 8001 and an item determination subunit 8003.

Specifically, the similarity calculation subunit 8001 may be configured to calculate a similarity between the first candidate item and each other item in the item set, and the item determination subunit 8003 may be configured to use, as the second candidate item, an item whose similarity in the item set is greater than a preset similarity threshold.

According to an exemplary embodiment of the present disclosure, referring to fig. 9, the acquisition characteristic determining module 25 may include a first acquisition characteristic determining unit 901.

Specifically, the first acquired feature determining unit 901 may be configured to perform determining a first feature of the target user according to an item interaction behavior of the target user within a first preset time period, determining a second feature of the target user based on a user portrait of the target user, extracting basic information from information of the target item and determining the first feature of the target item according to the basic information, extracting dynamic information from information of the target item and determining the second feature of the target item according to the dynamic information, and determining an item acquired feature of the target user by combining the first feature and the second feature of the target user and the first feature and the second feature of the target item.

According to an exemplary embodiment of the present disclosure, referring to fig. 10, the potential user determination device 10 may include a second model training module 101 in addition to the item sequence construction module 21, the intent item prediction module 23, the acquisition characteristic determination module 25, and the potential user determination module 27, as compared to the potential user determination device 2.

Specifically, the second model training module 101 may be configured to perform determining a plurality of second users, determining item acquisition characteristics of each of the second users, determining user information of presence intention behavior within a second preset time period, determining training samples of a probability model based on the item acquisition characteristics of each of the second users and the user information of presence intention behavior within the second preset time period, and training the probability model using the training samples of the probability model.

It should be appreciated that the second model training module 101 may also be included in the potential user determination device 5, and correspondingly the first model training module 51 may also be included in the potential user determination device 10.

According to an exemplary embodiment of the present disclosure, referring to fig. 11, the second model training module 101 may include a second acquisition characteristic determining unit 111.

Specifically, the second acquisition characteristic determining unit 111 may be configured to perform determining a first characteristic of the second user according to an item interaction behavior of the second user within a third preset period of time, determining a second characteristic of the second user based on a user image of the second user, extracting basic information from information of a predetermined item and determining the first characteristic of the predetermined item according to the basic information, extracting dynamic information from the information of the predetermined item and determining the second characteristic of the predetermined item according to the dynamic information, and combining the first and second characteristics of the second user and the first and second characteristics of the predetermined item to determine an item acquisition characteristic of the second user.

According to an exemplary embodiment of the present disclosure, referring to fig. 12, the potential user determination module 27 may include a potential user determination unit 121.

Specifically, the potential user determining unit 121 may be configured to determine whether the target item acquisition probability is greater than a preset probability threshold, and determine the target user as a potential user who acquires the target item if the target item acquisition probability is greater than the preset probability threshold.

According to an exemplary embodiment of the present disclosure, referring to fig. 13, the potential user determination device 13 may include a coupon information push module 131 in addition to the item sequence construction module 21, the intention item prediction module 23, the acquisition characteristic determination module 25, and the potential user determination module 27, as compared to the potential user determination device 2.

Specifically, the offer information pushing module 131 is configured to push first offer information to the target user if the target user has an intention achievement action for the target object within a fourth preset time period, and push second offer information to the target user if the target user does not have an intention achievement action for the target object within the fourth preset time period.

It should be noted that the offer information pushing module 131 may also be included in the potential user determining apparatus 5 or the potential user determining apparatus 10.

Since each functional module of the program execution performance analysis device according to the embodiment of the present invention is the same as that of the above-described method embodiment of the present invention, a detailed description thereof will be omitted.

In an exemplary embodiment of the present disclosure, a computer-readable storage medium having stored thereon a program product capable of implementing the method described above in the present specification is also provided. In some possible embodiments, the various aspects of the invention may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the steps according to the various exemplary embodiments of the invention as described in the "exemplary methods" section of this specification, when said program product is run on the terminal device.

Referring to fig. 14, a program product 1400 for implementing the above-described method according to an embodiment of the present invention is described, which may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of a readable storage medium include an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical disk, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the preceding.

The computer readable signal medium may include a data signal propagated in baseband or as part of a carrier wave with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

Those skilled in the art will appreciate that the various aspects of the invention may be implemented as a system, method, or program product. Accordingly, aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, micro-code, etc.) or an embodiment combining hardware and software aspects that may be referred to herein collectively as a "circuit," module "or" system.

An electronic device 1500 according to such an embodiment of the invention is described below with reference to fig. 15. The electronic device 1500 shown in fig. 15 is merely an example, and should not be construed as limiting the functionality and scope of use of embodiments of the present invention.

As shown in fig. 15, the electronic device 1500 is embodied in the form of a general purpose computing device. The components of the electronic device 1500 may include, but are not limited to, the at least one processing unit 1510, the at least one storage unit 1520, a bus 1530 connecting the different system components (including the storage unit 1520 and the processing unit 1510), and a display unit 1540.

Wherein the storage unit stores program code that is executable by the processing unit 1510 such that the processing unit 1510 performs steps according to various exemplary embodiments of the present invention described in the above section of the "exemplary method" of the present specification. For example, the processing unit 1510 may perform step S12 shown in FIG. 1, wherein a plurality of item information are determined according to item interaction behaviors of a target user, each item information is converted into an item vector, and a first item sequence composed of the plurality of item vectors is constructed, step S14, wherein the first item sequence is input into a trained intention item prediction model to determine a second item sequence, step S16, wherein a target item is determined based on the second item sequence and an item acquisition feature of the target user is determined according to the information of the target item, and step S18, wherein the item acquisition feature is input into a trained probability model to determine a target item acquisition probability of the target user and whether the target user is a potential user who acquires the target item according to the target item acquisition probability.

The storage unit 1520 may include readable media in the form of volatile memory units such as Random Access Memory (RAM) 15201 and/or cache memory 15202, and may further include Read Only Memory (ROM) 15203.

The storage unit 1520 may also include a program/utility 15204 having a set (at least one) of program modules 15205, such program modules 15205 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.

Bus 1530 may be a bus representing one or more of several types of bus structures including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

The electronic device 1500 may also communicate with one or more external devices 1600 (e.g., keyboard, pointing device, bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 1500, and/or any device (e.g., router, modem, etc.) that enables the electronic device 1500 to communicate with one or more other computing devices. Such communication may occur through an input/output (I/O) interface 1550. Also, the electronic device 1500 may also communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, for example, the Internet, through a network adapter 1560. As shown, the network adapter 1560 communicates with other modules of the electronic device 1500 over the bus 1530. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with electronic device 1500, including, but not limited to, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, data backup storage systems, and the like.

From the above description of embodiments, those skilled in the art will readily appreciate that the example embodiments described herein may be implemented in software, or may be implemented in software in combination with the necessary hardware. Thus, the technical solution according to the embodiments of the present disclosure may be embodied in the form of a software product, which may be stored in a non-volatile storage medium (may be a CD-ROM, a U-disk, a mobile hard disk, etc.) or on a network, including several instructions to cause a computing device (may be a personal computer, a server, a terminal device, or a network device, etc.) to perform the method according to the embodiments of the present disclosure.

Furthermore, the above-described drawings are only schematic illustrations of processes included in the method according to the exemplary embodiment of the present invention, and are not intended to be limiting. It will be readily appreciated that the processes shown in the above figures do not indicate or limit the temporal order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, among a plurality of modules.

It should be noted that although in the above detailed description several modules or units of a device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit in accordance with embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into a plurality of modules or units to be embodied.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.