User Life Stage Prediction Method, Device, Computer Equipment and Storage Medium

Abstract

The present invention discloses a method, apparatus, computer device, and storage medium for predicting a user's life stage. The method includes: determining a feature vector of a user to be predicted; respectively determining the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set, where different user sets belong to different life stages, the feature vector of a user set is determined according to the feature vectors of the sample users included in the user set, and the sample users included in the user set are determined according to the first behavior data of multiple sample users; determining the life stage to which the user set corresponding to the satisfied preset condition belongs as the life stage of the user to be predicted. In this method for predicting a user's life stage, the determined feature vector of the user set is relatively accurate, thereby achieving a relatively accurate prediction of the user's life stage.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of computers, and in particular, to a method, apparatus, computer device, and storage medium for predicting a user's life stage. Background Art

[0002] In the operation of an application (APP), as the number of APP users continues to increase, user operation becomes particularly important, especially the demand for refined operation has been greatly strengthened. At this time, it is necessary to stratify the overall users and divide the users into different life stages according to user characteristics to achieve refined and accurate operation.

[0003] Currently, according to the experience theory of relevant business experts or senior product managers, key user behaviors and key indicators can be extracted, and then based on the key behavior characteristics and the values of key indicators defined by combining historical experience and the actual user data distribution, the life stage of the user can be predicted.

[0004] However, in the above process, since the key behavior characteristics and the values of key indicators defined by combining historical experience and the actual user data distribution are not accurate, the accuracy of predicting the user's life stage is relatively low. Summary of the Invention

[0005] The present invention provides a method, apparatus, computer device, and storage medium for predicting a user's life stage to solve the technical problem of relatively low accuracy of the current method for predicting a user's life stage.

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting a user's life stage, including:

[0007] Determine the feature vector of the user to be predicted;

[0008] Respectively determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set; wherein, different user sets belong to different life stages, and the feature vector of the user set is determined according to the feature vectors of the sample users included in the user set, and the sample users included in the user set are determined according to the first behavior data of multiple sample users;

[0009] Determine the life stage to which the user set corresponding to the similarity meeting the preset condition belongs as the life stage of the user to be predicted.

[0010] In a second aspect, an embodiment of the present invention further provides a device for predicting a user's life stage, including:

[0011] A first determination module, configured to determine the feature vector of the user to be predicted;

[0012] A second determination module, configured to respectively determine the similarity between the feature vector of the to-be-predicted user and the feature vectors of each pre-determined user set; wherein, different user sets belong to different life stages, and the feature vector of the user set is determined according to the feature vectors of the sample users included in the user set, and the sample users included in the user set are determined according to the first behavior data of multiple sample users;

[0013] A third determination module, configured to determine the life stage to which the user set corresponding to the similarity meeting the preset condition belongs as the life stage of the to-be-predicted user.

[0014] In a third aspect, an embodiment of the present invention further provides a computer device, which includes:

[0015] One or more processors;

[0016] A memory, configured to store one or more programs;

[0017] When the one or more programs are executed by the one or more processors, the one or more processors implement the user life stage prediction method provided in the first aspect.

[0018] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the user life stage prediction method provided in the first aspect.

[0019] This embodiment provides a user life stage prediction method, device, computer device and storage medium. The method includes: determining the feature vector of the to-be-predicted user; respectively determining the similarity between the feature vector of the to-be-predicted user and the feature vectors of each pre-determined user set, wherein different user sets belong to different life stages, the feature vector of the user set is determined according to the feature vectors of the sample users included in the user set, and the sample users included in the user set are determined according to the first behavior data of multiple sample users; determining the life stage to which the user set corresponding to the similarity meeting the preset condition belongs as the life stage of the to-be-predicted user. In this user life stage prediction method, the feature vector of the user set is determined in advance by combining the first behavior data of the sample users and the feature vectors of the sample users, and then the life stage of the to-be-predicted user is determined based on the feature vector of the to-be-predicted user and the feature vector of the user set. Compared with the related art, the feature vector of the user set determined in this embodiment is relatively accurate, so that the life stage of the to-be-predicted user determined based on the feature vector of the user set and the feature vector of the to-be-predicted user is relatively accurate. Thus, a user life stage prediction with relatively high accuracy is achieved. Description of the Drawings

[0020] Figure 1 Schematic flowchart of an embodiment of a user life stage prediction method provided by the present invention;

[0021] Figure 2 Schematic flowchart in another embodiment of a user life stage prediction method provided by the present invention;

[0022] Figure 3 Schematic diagram of the process of determining the feature vector of the user set in the embodiment of the user life stage prediction method provided by the present invention;

[0023] Figure 4 Schematic flowchart of the process of determining the feature vector of each sample user in the embodiment of the user life stage prediction method provided by the present invention;

[0024] Figure 5 Schematic diagram of the process of determining the first sub-vector corresponding to the second behavior data of the sample user in the embodiment of the user life stage prediction method provided by the present invention;

[0025] Figure 6 Schematic flowchart in yet another embodiment of a user life stage prediction method provided by the present invention;

[0026] Figure 7 Schematic diagram of the process of determining the first sub-vector corresponding to the behavior data of the user to be predicted in the embodiment of the user life stage prediction method provided by the present invention;

[0027] Figure 8 Schematic structural diagram of a user life stage prediction device provided by the present invention;

[0028] Figure 9 Schematic structural diagram of another user life stage prediction device provided by the present invention;

[0029] Figure 10 Schematic structural diagram of a computer device provided by the present invention. Detailed implementation manners

[0030] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the present invention, rather than limiting the present invention. Additionally, it should be noted that for the sake of description, only parts related to the present invention are shown in the accompanying drawings rather than all the structures.

[0031] Figure 1Schematic flowchart of an embodiment of a method for predicting a user's life stage provided by the present invention. This embodiment is applicable to scenarios where the life stage of a user of an APP is predicted. This embodiment can be executed by a user life stage prediction device, which can be implemented in software and / or hardware, and can be integrated into a computer device. As Figure 1 shown, the user life stage prediction method provided in this embodiment includes the following steps:

[0032] Step 101: Determine the feature vector of the user to be predicted.

[0033] Specifically, the user to be predicted in this embodiment is a user of the APP. Exemplarily, the APP in this embodiment can be a parent user in a classroom management APP.

[0034] In one implementation, the feature vector of the user to be predicted can be determined according to the behavior data of the user to be predicted. Here, the behavior data can be the data of the behavior of the user to be predicted of a preset type in the APP, for example, a click operation on a certain page or a certain item. Here, the item can be a certain module in the page.

[0035] After obtaining the behavior data of the user to be predicted, the behavior data can be converted into a vector form, and this vector is the feature vector of the user to be predicted.

[0036] In another implementation, the feature vector of the user to be predicted can be determined according to the behavior data of the user to be predicted and the basic features of the user to be predicted. Here, the basic features of the user to be predicted can be the attribute information of the user. In the scenario where the user to be predicted is a parent user in a classroom management APP, the basic features of the user to be predicted can include at least one of the following: the school stage and grade of the student corresponding to the user to be predicted, the registration time of the user to be predicted, and the login-related features of the user to be predicted. Among them, the login-related features can be the login time, the type of login device, etc.

[0037] After obtaining the behavior data of the user to be predicted and the basic features of the user to be predicted: the behavior data can be converted into a vector form; the discrete features in the basic features can be encoded, for example, using one-hot encoding to digitize the discrete features, and the continuous variables in the basic features can be normalized, so as to convert the basic features into a vector form; the vector corresponding to the behavior data and the vector corresponding to the basic features are spliced to form the feature vector of the user to be predicted.

[0038] In another implementation manner, a feature vector of the user to be predicted can be determined according to the behavior data of the user to be predicted, the basic features of the user to be predicted, and the behavior data of the associated users of the user to be predicted. In the scenario where the user to be predicted is a parent user in a classroom management APP, the associated users of the user to be predicted can be teacher users. This implementation manner will be described in detail in the embodiments hereinafter.

[0039] Step 102: Determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set respectively.

[0040] Among them, different user sets belong to different life stages. The feature vector of a user set is determined according to the feature vectors of the sample users included in the user set. The sample users included in the user set are determined according to the first behavior data of multiple sample users.

[0041] Specifically, in this embodiment, before step 102, the sample users can be classified according to the first behavior data of the sample users and divided into multiple user sets. Then, for each user set, based on the feature vectors of the sample users included in the user set, the feature vector of the user set is pre-determined. The determination process of the feature vector of the sample users in this embodiment is similar to the process of determining the feature vector of the user to be predicted in step 101, and will not be elaborated here.

[0042] The user sets in this embodiment can have their own belonging life stages, and different user sets belong to different life stages.

[0043] Optionally, in this embodiment, the number of user sets is 4, and the life stages to which all user sets belong include: the novice period, the growth period, the mature period, and the decline period.

[0044] Exemplarily, the sample users in this embodiment can be parent users in a classroom management APP.

[0045] Optionally, the first behavior data of the sample users can include at least one of the following: consumption behavior data, registration behavior data, and login behavior data.

[0046] This way of determining the feature vector of the user set combines the first behavior data of the sample users and the feature vectors of the sample users, making the determined feature vector of the user set relatively accurate. Thus, the life stage of the user to be predicted determined based on the feature vector of the user set and the feature vector of the user to be predicted is relatively accurate, and the prediction accuracy is relatively high.

[0047] After the feature vectors of the user sets are pre-determined, in this embodiment, the similarity between the feature vector of the user to be predicted and the feature vectors of the user sets can be determined respectively.

[0048] In this embodiment, the following at least one similarity determination algorithm can be used to determine the similarity: inner product similarity, cosine similarity, Euclidean similarity, etc.

[0049] It can be understood that the number of user sets determines the corresponding number of similarities.

[0050] Step 103: Determine the life stage of the user set whose corresponding similarity meets the preset condition as the life stage of the user to be predicted.

[0051] Specifically, after determining the similarities between the feature vectors of the user to be predicted and the feature vectors of each user set respectively, the life stage of the user set whose corresponding similarity meets the preset condition can be determined as the life stage of the user to be predicted.

[0052] Optionally, the preset condition here can be the maximum similarity. Correspondingly, step 103 can be specifically: Determine the life stage corresponding to the user set with the maximum similarity as the life stage of the user to be predicted.

[0053] In this embodiment, after determining the life stage of the user to be predicted, the user to be predicted can be refined and precisely operated according to the life stage of the predicted user, so as to obtain the highest return on investment (ROI) with the lowest cost.

[0054] Optionally, after step 103, the user life stage prediction method provided in this embodiment may further include the following steps: Determine the operation message corresponding to the user to be predicted according to the mapping relationship between the life stage of the user and the operation message, and the life stage of the user to be predicted; Send the operation message corresponding to the user to be predicted to the user to be predicted. Among them, the operation message corresponding to the user to be predicted can be sent to the user to be predicted through an APP, text message, instant messaging software, email, etc.

[0055] The following uses a specific example to illustrate the user life stage prediction method provided in this embodiment. In this embodiment, four user sets are determined in advance: the novice stage user set (the life stage to which this set belongs is the novice stage), the growth stage user set (the life stage to which this set belongs is the growth stage), the mature stage user set (the life stage to which this set belongs is the mature stage), and the decline stage user set (the life stage to which this set belongs is the decline stage). Moreover, for the feature vectors of the sample users included in each user set, the feature vectors of the user sets are determined: the feature vector A of the novice stage user set, the feature vector B of the growth stage user set, the feature vector C of the mature stage user set, and the feature vector D of the decline stage user set. First, step 101 is executed to determine the feature vector of the user to be predicted; then, step 102 is executed to respectively determine the similarities between the feature vector of the user to be predicted and the feature vector A of the novice stage user set, the feature vector B of the growth stage user set, the feature vector C of the mature stage user set, and the feature vector D of the decline stage user set; then, step 103 is executed. Assume that the preset condition is the maximum similarity, and assume that the similarity between the feature vector of the user to be predicted and the feature vector B of the growth stage user set is the largest. Then, the life stage to which the growth stage user set belongs is determined as the life stage of the user to be predicted, that is, the life stage of the user to be predicted is the growth stage. In the above process, since the feature vectors of the user sets are determined by combining the first behavior data of the sample users and the feature vectors of the sample users, the determined feature vectors of the user sets are relatively accurate. Therefore, the life stage of the user to be predicted determined based on the feature vectors of the user sets and the feature vector of the user to be predicted is relatively accurate, and the prediction accuracy is relatively high.

[0056] This embodiment provides a user life stage prediction method, including: determining the feature vector of the user to be predicted; respectively determining the similarities between the feature vector of the user to be predicted and the feature vectors of each user set determined in advance, where different user sets belong to different life stages, and the feature vectors of the user sets are determined according to the feature vectors of the sample users included in the user sets, and the sample users included in the user sets are determined according to the first behavior data of multiple sample users; determining the life stage to which the user set corresponding to the satisfied preset condition of the similarity belongs as the life stage of the user to be predicted. In this user life stage prediction method, the feature vectors of the user sets are determined in advance by combining the first behavior data of the sample users and the feature vectors of the sample users, and then the life stage of the user to be predicted is determined based on the feature vector of the user to be predicted and the feature vectors of the user sets. Compared with the related art, the determined feature vectors of the user sets in this embodiment are relatively accurate, so that the life stage of the user to be predicted determined based on the feature vectors of the user sets and the feature vector of the user to be predicted is relatively accurate. Therefore, a user life stage prediction with relatively high accuracy is realized.

[0057] Figure 2 This is a schematic flowchart of another embodiment of the user life stage prediction method provided by the present invention. Based on the embodiments and various optional implementation manners shown below, a detailed description will be given to the process of determining the feature vectors of each user set. As Figure 1 shown, the user life stage prediction method provided by this embodiment includes the following steps: Figure 2 shown, the user life stage prediction method provided by this embodiment includes the following steps:

[0058] Step 201: Divide the sample users into multiple user sets according to the first behavior data of multiple sample users.

[0059] From the perspective of the human life stage, it is divided into infancy, youth, middle-aged and strong-aged, and old age. Humans in each life stage have completely different behavior patterns and characteristics every day, but the behavior patterns and characteristics in the same period are roughly the same. For example, when a person is in infancy, the behavior every day is basically crying - drinking milk - sleeping - crying - drinking milk - sleeping. When a person is in youth, it becomes getting up - going to school - attending classes - leaving school - doing homework. In middle-aged and strong-aged, it is getting up - commuting - working - overtime - entertainment every day. When entering old age, it becomes completely different, such as raising flowers - walking the bird - square dancing. Mapping the above ideas to user life stage prediction, when the number of users and behaviors are rich enough, it will be found that the behavior patterns of users in different life stages are also different, and the behaviors of users in the same life stage are also roughly similar.

[0060] In this embodiment, the sample users can be divided into multiple user sets by combining the first behavior data of multiple sample users.

[0061] Exemplarily, the first behavior data of the sample users in this embodiment includes at least one of the following data: consumption behavior data, registration behavior data, and login behavior data. Among them, the consumption behavior data may include: the time of the sample user's last transaction, the transaction frequency of the sample user, and the average single transaction amount of the sample user. The registration behavior data may include the duration from the registration time to the current time. The login behavior data may include the duration from the last login time to the current time and the login duration within the most recent preset duration (for example, the most recent K days), etc.

[0062] Optionally, the mapping relationships between the consumption behavior data, registration behavior data, and login behavior data of the sample users and the life stages of the sample users can be determined. Thereafter, for each sample user, according to the consumption behavior data, registration behavior data, and login behavior data of the sample user, and the mapping relationship, the life stage corresponding to the sample user is determined. The sample users with the same corresponding life stage are determined as a user set. Thus, the sample users are divided into multiple user sets.

[0063] For example, if the number of days since the registration of a certain sample user is less than the preset registration days threshold, the consumption of the sample user on the APP is less than the preset consumption threshold, or the duration from the time of the most recent consumption to the current time is less than the preset consumption duration threshold, or the consumption frequency is less than the preset frequency threshold, and the number of days since the most recent login of the sample user is less than the preset login duration threshold, it can be considered that the sample user is a typical user in the novice period.

[0064] Optionally, the sample users can be selected to ensure that the difference in the number of sample users included in each user set is within a preset range.

[0065] Step 202: Determine the feature vector of each sample user.

[0066] It should be noted that there is no chronological relationship between step 201 and step 202.

[0067] Figure 4 This is a schematic flowchart of determining the feature vector of each sample user in the embodiment of the user life stage prediction method provided by the present invention. As Figure 4 shown, step 202 may include the following steps.

[0068] Step 2021: According to the second behavior data of the sample user, determine the first sub-vector corresponding to the second behavior data of the sample user.

[0069] Exemplarily, the second behavior data of the sample user may include at least one of the following data: operation data of a preset page and operation data of a preset item. Here, the preset page may be the home page. In the scenario where the APP is a classroom management APP, the preset item may be courses, growth manuals, etc.

[0070] More specifically, the process of determining the first sub-vector corresponding to the second behavior data of the sample user can be as follows: taking the second behavior data of the sample user with a preset time interval as the time unit, forming a behavior sequence for each time unit; determining the high-dimensional vector mapped by the behavior sequence of each time unit; multiplying the high-dimensional vectors corresponding to the behavior sequences of all time units of the sample user by different time decay factors, and splicing them into the first sub-vector corresponding to the second behavior data of the sample user. Among them, the larger the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit closer to the current time, and the smaller the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit farther from the current time.

[0071] Exemplarily, the preset time interval can be one day.

[0072] Figure 5 This is a schematic diagram of the process of determining the first sub-vector corresponding to the second behavior data of the sample user in the embodiment of the user life stage prediction method provided by the present invention. As Figure 5 shown, taking the sample user 11 as an example for illustration. The second behavior data of the sample user 11 is obtained. Exemplarily, the second behavior data can be opening the home page, clicking on the course, opening the growth manual, etc. Figure 5 Among them, it is assumed that the second behavior data of the sample user 11 includes behavior data h1, behavior data h2,..., behavior data hh.

[0073] After that, taking the second behavior data of the sample user 11 with a preset time interval as the time unit, forming a behavior sequence for each time unit. In the scenario where the preset time interval can be one day, it can be in units of days to form a behavior sequence for each day. Figure 5 Among them, the behavior sequence of the time unit e1 includes: behavior data h1, behavior data h2,..., behavior data ht. The behavior sequence of the time unit e2 includes: behavior data h8, behavior data h9,..., behavior data hw...., the behavior sequence of the time unit ey includes: behavior data h7, behavior data h6,..., behavior data hx.

[0074] Then, determine the high-dimensional vector mapped by the behavior sequence of each time unit. Exemplarily, the Word2Vec method can be used to determine the high-dimensional vector mapped by the behavior sequence of each time unit. The Word2Vec method can take into account the context information and train the similarity of the behavior data to better represent the similarity of the sample user's behavior sequence, thereby improving the accuracy of the determined first sub-vector corresponding to the second behavior data of the sample user. Figure 5 Among them, the high-dimensional vector mapped by the behavior sequence of the time unit e1 is e11, the high-dimensional vector mapped by the behavior sequence of the time unit e2 is e12,..., and the high-dimensional vector mapped by the behavior sequence of the time unit ey is e1y.

[0075] After that, the high-dimensional vectors corresponding to the behavior sequences of all time units of the sample user 11 are multiplied by the corresponding time decay factors and concatenated into the first sub-vector corresponding to the second behavior data of the sample user. The time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit closer to the current time is larger, and the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit farther from the current time is smaller.

[0076] Optionally, when concatenating, it can be concatenated in the order of the time units before and after, that is, the high-dimensional vector mapped by the behavior sequence of the earlier time unit is located more forward in the first sub-vector. Of course, other rules can also be used for concatenation, as long as the concatenation rules of the first sub-vectors of all sample users are consistent.

[0077] The reason for the need for the time decay factor is as follows: The second behavior data of the sample user may span multiple life stages. In this embodiment, when calculating the first sub-vector corresponding to the second behavior data of a single sample user, the attention (ATTENTION) mechanism, that is, the time decay factor, is added. It can be understood that a greater weight is given to the newer (i.e., closer to the current time) behavior sequences, ensuring that the more recent behavior sequences are more important. In this way, when determining the first sub-vector corresponding to the second behavior data of the sample user, more attention can be paid to the most recent behavior sequences and the behavior sequences farther away can be ignored (because the behavior sequences farther away may be in the previous life stage). That is to say, considering the time decay factor in determining the first sub-vector corresponding to the second behavior data of the sample user can improve the accuracy of the determined first sub-vector corresponding to the second behavior data of the sample user, so as to ensure the accuracy of the feature vectors of the user set determined subsequently, and further ensure the accuracy of the user life stage prediction.

[0078] Step 2022: Determine the second sub-vector corresponding to the basic features of the sample user according to the basic features of the sample user.

[0079] The basic features of the sample user in this embodiment can be obtained from the user portrait of the sample user.

[0080] Exemplarily, in the scenario where the sample user is a parent user in the classroom management APP, the basic features of the sample user include at least one of the following: the school stage and grade of the student corresponding to the sample user, the registration time of the sample user, and the login-related features of the sample user.

[0081] In step 2022, the discrete features in the basic features can be encoded. For example, one-hot encoding can be used to digitize the discrete features, and the continuous variables in the basic features can be normalized. Thus, the basic features are converted into vector form to form a second sub-vector corresponding to the basic features of the sample user.

[0082] Exemplarily, the process of one-hot encoding discrete features can be as follows: For example, if the school stage feature has three values: primary school, junior high school, and senior high school, then the corresponding encoding vectors are [1 0 0], [0 1 0], and [0 0 1]. When normalizing continuous variables, the continuous variables can be normalized to the range between 0 and 1.

[0083] After each feature in the basic features is converted into digital form or normalized, they are concatenated to form a second sub-vector corresponding to the basic features of the sample user.

[0084] Optionally, during the concatenation process, different weights can be assigned to different features to improve the accuracy of the determined second sub-vector.

[0085] Step 2023: Determine a third sub-vector corresponding to the associated user of the sample user according to the behavior data of the associated user of the sample user.

[0086] Exemplarily, in the scenario where the APP is a classroom management APP and the sample user is a parent user in the classroom management APP, the associated user of the sample user can be a teacher user. The behavior data of the associated user of the sample user can include at least one of the following: the time when the teacher user sends a comment, the number of times the teacher user sends a comment to the sample user, and other data.

[0087] In step 2023, similar to step 2022, the discrete data in the behavior data of the associated user of the sample user is encoded. For example, one-hot encoding is used to digitize the discrete data, and the continuous data in the behavior data of the associated user of the sample user is normalized. Thus, the behavior data of the associated user of the sample user is converted into vector form to form a third sub-vector corresponding to the associated user of the sample user.

[0088] Step 2024: Concatenate the first sub-vector corresponding to the second behavior data of the sample user, the second sub-vector corresponding to the basic features of the sample user, and the third sub-vector corresponding to the associated user of the sample user according to a preset rule to form a feature vector of the sample user.

[0089] The preset rule here can include the weights of different sub-vectors during concatenation, the order of concatenation, etc. This preset rule can be set according to actual needs.

[0090] In the process of determining the feature vector of the sample user in this embodiment, the second behavior data of the sample user, the basic characteristics of the sample user and the behavior data of the sample user's associated users are comprehensively considered, so that the determined feature vector of the sample user is more comprehensive and accurate, which effectively improves the distinction between users at different life stages and further improves the accuracy of user life stage prediction.

[0091] Optionally, before the second behavior data of the sample user is used as a time unit to form a behavior sequence for each time unit, the method may further include: obtaining the second behavior data of the sample user in a preset time period before he no longer logs in. This implementation method is to ensure that each sample user has behavior data, so what is extracted is the second behavior data of the preset time period before the sample user no longer logs in, rather than the second behavior data of the preset time period before the current time, so as to improve the accuracy of the feature vector of the determined sample user.

[0092] In one implementation, the key core behavior data (i.e., the second behavior data) of the predefined sample users can be used to reduce invalid behavior noise in the user behavior log extraction, and the amount of calculation can be reduced to improve the calculation speed. In this embodiment, after the user behavior log is extracted, the user behavior log can be cleaned.

[0093] Step 203: For each user set, determine a feature vector of the user set according to feature vectors of sample users included in the user set.

[0094] Optionally, step 203 may include: clustering feature vectors of sample users included in the user set to obtain multiple sample user classes included in the user set; determining the feature vector of the user set based on the feature vector of the sample user included in the largest sample user class among the multiple sample user classes included in the user set.

[0095] In one implementation, this embodiment can use the currently available clustering algorithm to directly cluster the feature vectors of the sample users included in the user set, and clean out the feature vectors of some free sample users through the clustering results.

[0096] In another implementation, this embodiment may use currently available clustering algorithms and dimension reduction algorithms to reduce the dimension of feature vectors of sample users included in the user set and then perform clustering.

[0097] Exemplarily, in step 203, the average value of the feature vectors of the sample users included in the largest sample user class may be determined as the feature vector of the user set.

[0098] The above steps 201 to 203 are described below with a specific example. Figure 3This is a schematic diagram of the process of determining the feature vectors of user sets in the embodiment of the user life stage prediction method provided by the present invention. After step 201 is executed, as Figure 3 shown, four user sets are formed: the novice period user set, the growth period user set, the mature period user set, and the decline period user set. Among them, the novice period user set includes: sample user 11, sample user 12,..., sample user m. The growth period user set includes: sample user 21, sample user 22,..., sample user n. The mature period user set includes: sample user 31, sample user 32,..., sample user r. The decline period user set includes: sample user 41, sample user 42,..., sample user t. After step 202 is executed, the feature vectors of each sample user are determined. Then, step 203 is executed: according to the feature vectors of the sample users included in the novice period user set, determine the feature vector of the novice period user set; according to the feature vectors of the sample users included in the growth period user set, determine the feature vector of the growth period user set; according to the feature vectors of the sample users included in the mature period user set, determine the feature vector of the mature period user set; according to the feature vectors of the sample users included in the decline period user set, determine the feature vector of the decline period user set.

[0099] Step 204: Determine the feature vector of the user to be predicted.

[0100] Step 205: Respectively determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set.

[0101] Among them, the life stages of different user sets are different. The feature vector of a user set is determined according to the feature vectors of the sample users included in the user set. The sample users included in the user set are determined according to the first behavior data of multiple sample users.

[0102] Step 206: Determine the life stage of the user to be predicted as the life stage of the user set whose corresponding similarity meets the preset condition.

[0103] The implementation processes and technical principles of step 204 and step 101, step 205 and step 102, and step 206 and step 103 are similar and will not be elaborated here.

[0104] The user life stage prediction method provided in this embodiment first forms multiple user sets based on the first behavior data of sample users. Then, according to the feature vectors of the sample users included in the user set, the feature vector of the user set is determined, realizing the annotation of the user set based on the first behavior data of the sample users. Then, based on the feature vectors of the sample users, the feature vector of the user set is determined. The above process combines the first behavior data of the sample users and the feature vectors of the sample users to determine the feature vector of the user set. Compared with the related technology, the feature vector of the user set determined in this embodiment is more accurate, making the life stage of the to-be-predicted user determined based on the feature vector of this user set and the feature vector of the to-be-predicted user more accurate. Thus, a user life stage prediction with higher accuracy is realized.

[0105] Figure 6 It is a schematic flowchart in another embodiment of the user life stage prediction method provided by the present invention. The user life stage prediction method provided in this embodiment is based on the embodiments shown in Figure 1 the embodiments shown, Figure 2 the embodiments shown and various optional implementation manners, and makes a detailed description of the process of how to determine the feature vector of the to-be-predicted user. As Figure 6 shown, the user life stage prediction method provided in this embodiment includes the following steps:

[0106] Step 601: Determine the first sub-vector corresponding to the behavior data of the to-be-predicted user according to the behavior data of the to-be-predicted user.

[0107] Exemplarily, the behavior data of the to-be-predicted user here may be the second behavior data of the to-be-predicted user. The second behavior data of the to-be-predicted user may include at least one of the following data: operation data of a preset page and operation data of a preset item. Here, the preset page may be the home page. In the scenario where the APP is a classroom management APP, the preset item may be a course, a growth manual, etc.

[0108] It should be noted that the first behavior data of the to-be-predicted user may include at least one of the following data: consumption behavior data, registration behavior data, and login behavior data.

[0109] In one scenario, the to-be-predicted user may not have the first behavior data. The user life stage prediction method provided in this embodiment can realize predicting the life stage of the to-be-predicted user based on the second behavior data of the to-be-predicted user.

[0110] More specifically, the process of determining the first sub-vector corresponding to the behavior data of the user to be predicted may be as follows: taking the behavior data of the user to be predicted with a preset time interval as the time unit, forming a behavior sequence for each time unit; determining the high-dimensional vector mapped by the behavior sequence of each time unit; multiplying the high-dimensional vectors corresponding to the behavior sequences of all time units of the user to be predicted by different time decay factors, and splicing them into the first sub-vector corresponding to the behavior data of the user to be predicted. Among them, the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit closer to the current time is larger, and the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit farther from the current time is smaller.

[0111] Figure 7 FIG. is a schematic diagram of the process of determining the first sub-vector corresponding to the behavior data of the user to be predicted in the embodiment of the user life stage prediction method provided by the present invention. As Figure 7 shown, the behavior data of the user to be predicted is obtained. Exemplarily, the behavior data may be opening the home page, clicking on a course, opening a growth manual, etc. Figure 7 In, it is assumed that the behavior data of the user to be predicted includes behavior data d1, behavior data d2,..., behavior data dh.

[0112] After that, taking the behavior data of the user to be predicted with a preset time interval as the time unit, forming a behavior sequence for each time unit. In the scenario where the preset time interval can be days, it can be in days as the unit to form a daily behavior sequence. Figure 7 In, the behavior sequence of time unit j1 includes: behavior data d1, behavior data d2,..., behavior data dt. The behavior sequence of time unit j2 includes: behavior data d8, behavior data d9,..., behavior data dw...., the behavior sequence of time unit jy includes: behavior data d7, behavior data d6,..., behavior data dx.

[0113] Then, determine the high-dimensional vector mapped by the behavior sequence of each time unit. Exemplarily, the Word2Vec method can be used to determine the high-dimensional vector mapped by the behavior sequence of each time unit. The Word2Vec method can take into account the context information and train the similarity of the behavior data to better represent the similarity of the behavior sequence of the user to be predicted, thereby improving the accuracy of the determined first sub-vector corresponding to the behavior data of the user to be predicted. Figure 7 In, the high-dimensional vector mapped by the behavior sequence of time unit j1 is j11, the high-dimensional vector mapped by the behavior sequence of time unit j2 is j12,..., the high-dimensional vector mapped by the behavior sequence of time unit jy is j1y.

[0114] After that, multiply the high-dimensional vectors corresponding to the behavior sequences of all time units of the user to be predicted by different time decay factors, and splice them into a first sub-vector corresponding to the behavior data of the user to be predicted. The time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit closer to the current time is larger, and the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit farther from the current time is smaller.

[0115] Considering the time decay factor in the first sub-vector corresponding to the behavior data of the user to be predicted can improve the accuracy of the first sub-vector corresponding to the behavior data of the user to be predicted, and further ensure the accuracy of user life stage prediction.

[0116] Step 602: Determine a second sub-vector corresponding to the basic features of the user to be predicted according to the basic features of the user to be predicted.

[0117] The basic features of the user to be predicted in this embodiment can be obtained from the user portrait of the user to be predicted.

[0118] In step 602, the discrete features in the basic features of the user to be predicted can be encoded to digitalize the discrete features; the continuous variables in the basic features can be normalized. Thus, the basic features are converted into vector form to form a second sub-vector corresponding to the basic features of the user to be predicted.

[0119] After each feature in the basic features is converted into digital form or normalized, they are spliced into a second sub-vector corresponding to the basic features of the user to be predicted.

[0120] Optionally, different weights can be assigned to different features during the splicing process to improve the accuracy of the determined second sub-vector.

[0121] Step 603: Determine a third sub-vector corresponding to the associated user of the user to be predicted according to the behavior data of the associated user of the user to be predicted.

[0122] Exemplarily, in the scenario where the APP is a classroom management APP and the user to be predicted is a parent user in the classroom management APP, the associated user of the user to be predicted can be a teacher user. The behavior data of the associated user of the user to be predicted may include at least one of the following: the time when the teacher user sends a comment, the number of times the teacher user sends a comment to the sample user, and other data.

[0123] In step 603, similar to step 602, the discrete data in the behavioral data of the associated users of the user to be predicted is encoded to digitalize the discrete data, and the continuous data in the behavioral data of the associated users of the user to be predicted is normalized. Thus, the behavioral data of the associated users of the user to be predicted is converted into a vector form to form a third sub-vector corresponding to the associated users of the user to be predicted.

[0124] Step 604: According to a preset rule, concatenate the first sub-vector corresponding to the behavioral data of the user to be predicted, the second sub-vector corresponding to the basic features of the user to be predicted, and the third sub-vector corresponding to the associated users of the user to be predicted to form a feature vector of the user to be predicted.

[0125] The preset rule here may include the weights of different sub-vectors during concatenation, the order of concatenation, etc. The preset rule can be set according to actual requirements.

[0126] In the process of determining the feature vector of the user to be predicted in this embodiment, the behavioral data of the user to be predicted, the basic features of the user to be predicted, and the behavioral data of the associated users of the user to be predicted are comprehensively considered, so that the determined feature vector of the user to be predicted is more comprehensive and accurate, further improving the accuracy of user life stage prediction.

[0127] Step 605: Determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set respectively.

[0128] Among them, different user sets belong to different life stages. The feature vector of a user set is determined according to the feature vectors of the sample users included in the user set. The sample users included in the user set are determined according to the first behavioral data of multiple sample users.

[0129] Step 606: Determine the life stage to which the user set belongs, for which the corresponding similarity meets the preset condition, as the life stage of the user to be predicted.

[0130] The implementation processes and technical principles of step 605 and step 102, step 606 and step 103 are similar and will not be elaborated here.

[0131] In the method for predicting a user's life stage provided in this embodiment, the process of determining the feature vector of the user to be predicted includes: determining a first sub-vector corresponding to the behavior data of the user to be predicted according to the behavior data of the user to be predicted; determining a second sub-vector corresponding to the basic features of the user to be predicted according to the basic features of the user to be predicted; determining a third sub-vector corresponding to the associated user of the user to be predicted according to the behavior data of the associated user of the user to be predicted; and combining the first sub-vector corresponding to the behavior data of the user to be predicted, the second sub-vector corresponding to the basic features of the user to be predicted, and the third sub-vector corresponding to the associated user of the user to be predicted according to a preset rule to form the feature vector of the user to be predicted. In the method for predicting a user's life stage provided in this embodiment, when determining the feature vector of the user to be predicted, the behavior data of the user to be predicted, the basic features of the user to be predicted, and the behavior data of the associated user of the user to be predicted are comprehensively considered, so that the determined feature vector of the user to be predicted is relatively comprehensive and accurate, and further improves the accuracy of predicting the user's life stage.

[0132] Figure 8 It is a schematic structural diagram of a device for predicting a user's life stage provided by the present invention. The device for predicting a user's life stage provided in this embodiment includes the following modules: a first determination module 81, a second determination module 82, and a third determination module 83.

[0133] The first determination module 81 is used to determine the feature vector of the user to be predicted.

[0134] Optionally, the first determination module 81 is specifically configured to: determine a first sub-vector corresponding to the behavior data of the user to be predicted according to the behavior data of the user to be predicted; determine a second sub-vector corresponding to the basic features of the user to be predicted according to the basic features of the user to be predicted; determine a third sub-vector corresponding to the associated user of the user to be predicted according to the behavior data of the associated user of the user to be predicted; and combine the first sub-vector corresponding to the behavior data of the user to be predicted, the second sub-vector corresponding to the basic features of the user to be predicted, and the third sub-vector corresponding to the associated user of the user to be predicted according to a preset rule to form the feature vector of the user to be predicted.

[0135] The second determination module 82 is used to determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set respectively.

[0136] Among them, different user sets belong to different life stages. The feature vector of a user set is determined according to the feature vectors of the sample users included in the user set. The sample users included in the user set are determined according to the first behavior data of multiple sample users.

[0137] Optionally, in this embodiment, the number of user sets is 4. The life stages to which all user sets belong include: the novice stage, the growth stage, the mature stage, and the decline stage.

[0138] The third determination module 83 is configured to determine the life stage to which the user set corresponding to the corresponding similarity meets the preset condition as the life stage of the user to be predicted.

[0139] Optionally, the third determination module 83 is specifically configured to: determine the life stage corresponding to the user set with the largest corresponding similarity as the life stage of the user to be predicted.

[0140] Further, the device may include: a fourth determination module and a sending module. The fourth determination module is configured to determine the operation message corresponding to the user to be predicted according to the mapping relationship between the life stage of the user and the operation message, and the life stage of the user to be predicted. The sending module is configured to send the operation message corresponding to the user to be predicted to the user to be predicted.

[0141] Optionally, both the sample user and the user to be predicted are parent users in a classroom management application. The associated user of the user to be predicted is a teacher user.

[0142] Optionally, the basic features of the user to be predicted include at least one of the following: the school stage and grade of the student corresponding to the user to be predicted, the registration time of the user to be predicted, and the login-related features of the user to be predicted.

[0143] Optionally, the behavior data of the user to be predicted includes at least one of the following data: the operation data of a preset page and the operation data of a preset item.

[0144] The user life stage prediction device provided by the embodiments of the present invention can execute the user life stage prediction method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0145] Figure 9 It is a schematic structural diagram of another user life stage prediction device provided by the present invention. This embodiment is based on Figure 8 the shown embodiment and various optional implementation manners, and makes a detailed description of other modules of the user life stage prediction device. As Figure 9 shown, the user life stage prediction device provided by this embodiment further includes the following modules: a division module 91, a fifth determination module 92, and a sixth determination module 93.

[0146] The division module 91 is configured to divide the sample users into multiple user sets according to the first behavior data of the multiple sample users.

[0147] Optionally, the first-line behavior data of the sample user includes at least one of the following data: consumption behavior data, registration behavior data, and login behavior data.

[0148] The fifth determination module 92 is configured to determine the feature vector of each sample user.

[0149] Optionally, the fifth determination module 92 is specifically configured to: determine a first sub-vector corresponding to the second-line behavior data of the sample user according to the second-line behavior data of the sample user; determine a second sub-vector corresponding to the basic features of the sample user according to the basic features of the sample user; determine a third sub-vector corresponding to the associated user of the sample user according to the behavior data of the associated user of the sample user; and splice the first sub-vector corresponding to the second-line behavior data of the sample user, the second sub-vector corresponding to the basic features of the sample user, and the third sub-vector corresponding to the associated user of the sample user according to a preset rule to form the feature vector of the sample user.

[0150] In terms of determining the first sub-vector corresponding to the second-line behavior data of the sample user according to the second-line behavior data of the sample user, the fifth determination module 92 is specifically configured to: form a behavior sequence for each time unit with the second-line behavior data of the sample user using the preset time interval as the time unit; determine the high-dimensional vector mapped by the behavior sequence of each time unit; multiply the high-dimensional vectors corresponding to the behavior sequences of all time units of the sample user by different time decay factors and splice them into the first sub-vector corresponding to the second-line behavior data of the sample user; wherein, the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit closer to the current time is larger, and the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit farther from the current time is smaller.

[0151] In one implementation manner, the device provided in this embodiment further includes an acquisition module, configured to acquire the second-line behavior data of the sample user in a preset time period before the sample user stops logging in.

[0152] The sixth determination module 93 is configured to determine the feature vector of each user set according to the feature vectors of the sample users included in the user set.

[0153] Optionally, the sixth determination module 93 is specifically configured to: cluster the feature vectors of the sample users included in the user set to obtain multiple sample user classes included in the user set; and determine the feature vector of the user set according to the feature vectors of the sample users included in the largest sample user class among the multiple sample user classes included in the user set.

[0154] The user life stage prediction device provided in the embodiments of the present invention can execute the user life stage prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0155] Figure 10 The structural schematic diagram of a computer device provided by the present invention. As Figure 10 shown, the computer device includes a processor 110 and a memory 111. The number of processors 110 in the computer device may be one or more, Figure 10 and one processor 110 is taken as an example herein; the processor 110 and the memory 111 of the computer device may be connected by a bus or other means, Figure 10 and connection by a bus is taken as an example herein.

[0156] As a computer-readable storage medium, the memory 111 can be used to store software programs, computer-executable programs, and modules, such as program instructions and modules corresponding to the user life stage prediction method in the embodiments of the present invention (for example, the first determination module 81, the second determination module 82, and the third determination module 83 in the user life stage prediction device). By running the software programs, instructions, and modules stored in the memory 111, the processor 110 executes various functional applications and data processing of the computer device, that is, implements the above-mentioned user life stage prediction method.

[0157] The memory 111 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the computer device. In addition, the memory 111 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory 111 may further include a memory remotely provided relative to the processor 110, and these remote memories can be connected to the computer device through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0158] The present invention also provides a storage medium containing computer-executable instructions, and the computer-executable instructions are used to execute a user life stage prediction method when executed by a computer processor. The method includes:

[0159] Determine the feature vector of the user to be predicted;

[0160] Respectively determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set; wherein, different user sets belong to different life stages, and the feature vector of the user set is determined according to the feature vectors of the sample users included in the user set, and the sample users included in the user set are determined according to the first behavior data of multiple sample users;

[0161] Determine the life stage to which the user set corresponding to the similarity meeting the preset condition belongs as the life stage of the to-be-predicted user.

[0162] Certainly, for a storage medium containing computer-executable instructions provided by an embodiment of the present invention, the computer-executable instructions are not limited to the method operations as described above, and can also execute related operations in the user life stage prediction method provided by any embodiment of the present invention.

[0163] From the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general-purpose hardware. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as a floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk, or optical disc of a computer, etc., including several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0164] It should be noted that in the embodiments of the above user life stage prediction device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0165] Note that the above is only the preferred embodiment of the present invention and the applied technical principle. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described here, and various obvious changes, re-adjustments, and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for predicting a user's life stage, characterized in that, Including: Determine the feature vector of the user to be predicted, including: determining the feature vector of the user to be predicted according to the behavior data of the user to be predicted and the basic features of the user to be predicted, where the behavior data is the data of the preset type of behavior of the user to be predicted in the application program, and the basic features are the attribute information of the user to be predicted; Respectively determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set; wherein, different user sets belong to different life stages, and the feature vector of the user set is determined according to the feature vectors of the sample users included in the user set, and the sample users included in the user set are determined according to the first behavior data of multiple sample users; Determine the life stage to which the user set to which the corresponding similarity meets the preset condition belongs as the life stage of the user to be predicted; The number of the user sets is 4, and the life stages to which all the user sets belong include: the novice period, the growth period, the mature period, and the decline period; Before respectively determining the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set, the method further includes: Dividing the sample users into multiple user sets according to the first behavior data of the multiple sample users; Determine the feature vector of each sample user; For each user set, determine the feature vector of the user set according to the feature vectors of the sample users included in the user set; The first behavior data of the sample users includes at least one of the following data: consumption behavior data, registration behavior data, and login behavior data; The determining the feature vector of each sample user includes: Determine the first sub-vector corresponding to the second behavior data of the sample user according to the second behavior data of the sample user; Determine the second sub-vector corresponding to the basic features of the sample user according to the basic features of the sample user; Determine the third sub-vector corresponding to the associated user of the sample user according to the behavior data of the associated user of the sample user; Splice the first sub-vector corresponding to the second behavior data of the sample user, the second sub-vector corresponding to the basic features of the sample user, and the third sub-vector corresponding to the associated user of the sample user according to a preset rule to form the feature vector of the sample user; The determining the first sub-vector corresponding to the second behavior data of the sample user according to the second behavior data of the sample user includes: Taking the second behavior data of the sample user as a time unit at a preset time interval to form a behavior sequence for each time unit; Determine the high-dimensional vector mapped by the behavior sequence of each time unit; wherein, the high-dimensional vector is determined by using the Word2Vec method; Multiply the high-dimensional vectors corresponding to the behavior sequences of all time units of the sample user by different time decay factors, and splice them into the first sub-vector corresponding to the second behavior data of the sample user; among them, the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit closer to the current time is larger, and the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit farther from the current time is smaller.

2. The method according to claim 1, characterized in that Before forming the behavior sequence of each time unit with the second behavior data of the sample user as the time unit at a preset time interval, the method further includes: Obtain the second behavior data of the sample user in a preset time period before the sample user stops logging in.

3. The method according to claim 1, characterized in that, The determining the feature vector of the user set according to the feature vectors of the sample users included in the user set includes: Cluster the feature vectors of the sample users included in the user set to obtain multiple sample user classes included in the user set; Determine the feature vector of the user set according to the feature vectors of the sample users included in the largest sample user class among the multiple sample user classes included in the user set.

4. The method according to any one of claims 1 to 3, characterized in that, The determining the feature vector of the user to be predicted further includes: Determine the first sub-vector corresponding to the behavior data of the user to be predicted according to the behavior data of the user to be predicted; Determine the second sub-vector corresponding to the basic features of the user to be predicted according to the basic features of the user to be predicted; Determine the third sub-vector corresponding to the associated user of the user to be predicted according to the behavior data of the associated user of the user to be predicted; Splice the first sub-vector corresponding to the behavior data of the user to be predicted, the second sub-vector corresponding to the basic features of the user to be predicted, and the third sub-vector corresponding to the associated user of the user to be predicted according to a preset rule to form the feature vector of the user to be predicted.

5. The method according to any one of claims 1 to 3, characterized in that The determining the life stage of the user to be predicted as the life stage of the user set whose corresponding similarity meets the preset condition includes: Determine the life stage of the user to be predicted as the life stage of the user set with the largest corresponding similarity.

6. The method according to any one of claims 1 to 3, characterized in that, After determining the life stage of the user to be predicted as the life stage of the user set whose corresponding similarity meets the preset condition, the method includes: Determine the operation message corresponding to the user to be predicted according to the mapping relationship between the life stage of the user and the operation message, and the life stage of the user to be predicted; Send the operation message corresponding to the user to be predicted to the user to be predicted.

7. The method according to claim 4, wherein Both the sample user and the user to be predicted are parent users in a classroom management application program, and the associated user of the user to be predicted is a teacher user.

8. The method according to claim 7, wherein The basic features of the user to be predicted include at least one of the following: the school stage and grade of the student corresponding to the user to be predicted, the registration time of the user to be predicted, and the login-related features of the user to be predicted.

9. The method according to claim 4, characterized in that The behavior data of the user to be predicted includes at least one of the following data: operation data of a preset page and operation data of a preset item.

10. A user life stage prediction device, characterized in that, Include: The first determination module is used to determine the feature vector of the user to be predicted, including: determining the feature vector of the user to be predicted according to the behavior data of the user to be predicted and the basic features of the user to be predicted, where the behavior data is the data of the behavior of the preset type in the application program by the user to be predicted, and the basic features are the attribute information of the user to be predicted; The second determination module is used to determine the similarity between the feature vector of the user to be predicted and the feature vectors of each pre-determined user set respectively; where different user sets belong to different life stages, and the feature vector of the user set is determined according to the feature vectors of the sample users included in the user set, and the sample users included in the user set are determined according to the first behavior data of multiple sample users; The third determination module is used to determine the life stage to which the user set to which the corresponding similarity meets the preset condition belongs as the life stage of the user to be predicted; The number of the user sets is 4, and the life stages to which all the user sets belong include: the novice period, the growth period, the mature period, and the decline period; The device further includes: The division module is used to divide the sample users into multiple user sets according to the first behavior data of multiple sample users; The fifth determination module is used to determine the feature vector of each sample user; The sixth determination module is used to, for each user set, determine the feature vector of the user set according to the feature vectors of the sample users included in the user set; The first behavior data of the sample users includes at least one of the following data: consumption behavior data, registration behavior data, and login behavior data; The fifth determination module is specifically used to: determine the first sub-vector corresponding to the second behavior data of the sample user according to the second behavior data of the sample user; determine the second sub-vector corresponding to the basic features of the sample user according to the basic features of the sample user; determine the third sub-vector corresponding to the associated user of the sample user according to the behavior data of the associated user of the sample user; and splice the first sub-vector corresponding to the second behavior data of the sample user, the second sub-vector corresponding to the basic features of the sample user, and the third sub-vector corresponding to the associated user of the sample user according to a preset rule to form the feature vector of the sample user; In terms of determining the first sub-vector corresponding to the second behavior data of the sample user according to the second behavior data of the sample user, the fifth determination module is specifically used to: form a behavior sequence for each time unit with the second behavior data of the sample user as the time unit at a preset time interval; determine the high-dimensional vector mapped by the behavior sequence of each time unit; where the high-dimensional vector is determined by using the Word2Vec method; multiply the high-dimensional vectors corresponding to the behavior sequences of all time units of the sample user by different time decay factors and splice them into the first sub-vector corresponding to the second behavior data of the sample user; where the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit closer to the current time is larger, and the time decay factor multiplied by the high-dimensional vector mapped by the behavior sequence of the time unit farther from the current time is smaller.

11. A computer device, characterized in that, The computer device includes: One or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the user life stage prediction method according to any one of claims 1 to 9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the user life stage prediction method according to any one of claims 1 to 9.

Images