Method and apparatus for constructing user behavior prediction model based on deep recommendation model

Abstract

The application discloses a user behavior prediction model construction method, a user behavior prediction method, a user behavior prediction device, equipment and a medium. The user behavior prediction model construction method comprises the following steps: generating an original sample set according to historical behavior data of multiple users; generating a matrix corresponding to each original sample according to a behavior set in each original sample; forming a training sample set; training a deep recommendation model using the training sample set to obtain a user behavior prediction model. The user behavior prediction method comprises the following steps: obtaining user feature information of a to-be-predicted user and item feature information of a to-be-predicted item; obtaining a behavior prediction probability matrix of the to-be-predicted user to the to-be-predicted item; and verifying whether the to-be-predicted item is recommended to the to-be-predicted user according to the behavior prediction probability matrix. The technical scheme of the application provides a new way of constructing a user behavior prediction model to improve the accuracy of user behavior prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of model construction, in particular to a user behavior prediction model construction method, a user behavior prediction method, device, equipment and medium. BACKGROUND

[0002] With the rapid development of the Internet, information data in the network is growing exponentially, leading to information redundancy, which raises a problem, that is, how users can obtain information of their own interest in an effective time. At this time, the recommendation system based on user behavior prediction emerges as the times require, which greatly improves the user experience and effectively reduces the time consumption of users searching for information of their own interest.

[0003] The above recommendation system has two implementation methods: method 1 is to change multiple training targets into one target according to weights to make the model fit; method 2 is to construct multiple network structures, so that different networks are responsible for different training targets, and finally multiple training is fused to maximize the benefits of multiple targets through a certain measurement method.

[0004] The inventor found that the prior art has the following problems in the process of implementing the present application: method 1 completely separates each training target, which destroys the integrity of information and further affects the accuracy of recommendation; method 2 has too many human factors, which affects the recommendation results. SUMMARY

[0005] The present application provides a user behavior prediction model construction, user behavior prediction method, device, equipment and medium, which provides a new way to construct a user behavior prediction model and improves the accuracy of user behavior prediction.

[0006] In a first aspect, the present application provides a user behavior prediction model construction method, which comprises:

[0007] According to the historical behavior data of multiple users, an original sample set is generated, which includes user feature information of a set user, item feature information of a set item and a behavior set performed by the set user on the set item, and each behavior in the behavior set has a progressive relationship;

[0008] According to the behavior set in each original sample, a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample are generated;

[0009] According to user feature information of a set user in an original sample, item feature information of a set item, a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix, a training sample set is formed, and the second behavior feature matrix and the third behavior feature matrix are used as labeled data in the training sample;

[0010] The depth recommendation model is trained using the training sample set, and a user behavior prediction model is obtained.

[0011] In a second aspect, an embodiment of the present application provides a user behavior prediction method, which comprises:

[0012] Obtaining user feature information of a to-be-predicted user and item feature information of a to-be-predicted item;

[0013] Inputting the user feature information of the to-be-predicted user and the item feature information of the to-be-predicted item into a user behavior prediction model trained by the method for constructing the user behavior prediction model according to any embodiment of the present application, and obtaining a behavior prediction probability matrix of the to-be-predicted user on the to-be-predicted item;

[0014] According to the behavior prediction probability matrix, verifying whether the to-be-predicted item is recommended to the to-be-predicted user.

[0015] In a third aspect, an embodiment of the present application provides a construction device of a user behavior prediction model, which comprises:

[0016] An original sample set generation module is configured to generate an original sample set according to historical behavior data of multiple users, wherein the original sample includes user feature information of a set user, item feature information of a set item and a set of behaviors performed by the set user on the set item, and there is a progressive relationship between behaviors in the set of behaviors;

[0017] A feature matrix generation module is configured to generate a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample, respectively, according to the set of behaviors in each original sample;

[0018] A training sample set generation module is configured to form a training sample set according to the user feature information of the set user, the item feature information of the set item, the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix in the original sample, and the second behavior feature matrix and the third behavior feature matrix are used as labeled data in the training sample;

[0019] A model training module is configured to train a depth recommendation model using the training sample set, and obtain a user behavior prediction model.

[0020] In a fourth aspect, an embodiment of the present application provides a user behavior prediction device, which comprises:

[0021] characteristics information acquisition module, configured to acquire user characteristics information of a to-be-predicted user and item characteristics information of a to-be-predicted item;

[0022] a probability matrix acquisition module, configured to input the user characteristics information of the to-be-predicted user and the item characteristics information of the to-be-predicted item into a user behavior prediction model trained by the method for constructing a user behavior prediction model according to any one of the embodiments of the present application, and acquire a behavior prediction probability matrix of the to-be-predicted user on the to-be-predicted item.

[0023] a recommendation verification module, configured to verify whether to recommend the to-be-predicted item to the to-be-predicted user according to the behavior prediction probability matrix.

[0024] In a fifth aspect, an embodiment of the present application provides an electronic device, which comprises:

[0025] at least one processor; and

[0026] a memory connected with the at least one processor in communication; wherein

[0027] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the method for constructing a user behavior prediction model according to any one of the embodiments of the present application, or execute the user behavior prediction method according to any one of the embodiments of the present application.

[0028] In a sixth aspect, an embodiment of the present application provides a computer readable storage medium, which stores computer instructions, and the computer instructions are used to enable a processor to implement the method for constructing a user behavior prediction model according to any one of the embodiments of the present application, or implement the user behavior prediction method according to any one of the embodiments of the present application.

[0029] The technical scheme of the embodiment of the present application provides a new way of constructing a user behavior prediction model by generating an original sample set according to historical behavior data of multiple users, generating a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample respectively according to a behavior set in each original sample, forming a training sample set and labeled data according to user characteristics information of a set user, item characteristics information of a set item, the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix in the original sample, and finally training a deep recommendation model using the training sample set to obtain a user behavior prediction model. Based on the new user behavior prediction model, the accuracy of user behavior prediction can be effectively improved, the hit rate of recommended items on actual user demand can be improved, and the time consumption of users searching for interested information by themselves can be effectively reduced.

[0030] It should be understood that the matters described in this section are not intended to identify key or critical elements of the embodiments of the present application or to delineate the scope of the present application. Other features of the present application will be apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0032] Figure 1 is a flow chart of a user behavior prediction model construction method according to an embodiment of the present application;

[0033] Figure 2a is a flow chart of a user behavior prediction model construction method according to an embodiment of the present application;

[0034] Figure 2b is a framework diagram of a user behavior prediction model obtained by the method according to an embodiment of the present application;

[0035] Figure 3 is a flow chart of a user behavior prediction method according to an embodiment of the present application;

[0036] Figure 4 is a structural schematic diagram of a user behavior prediction model construction device according to an embodiment of the present application;

[0037] Figure 5 is a structural schematic diagram of a user behavior prediction device according to an embodiment of the present application;

[0038] Figure 6 is a structural schematic diagram of an electronic device implementing the user behavior prediction model construction method and the user behavior prediction method according to an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the technical personnel in the art better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0040] It is to be understood that the terminology "first", "second" and the like used in the specification and the claims of the application as well as the preceding description of the drawings is merely used to distinguish similar objects and does not necessarily imply a specific order or chronology. It is to be understood that the data thus used can be interchanged, where appropriate, so that the embodiments of the application described herein can be carried out in a different order than the one illustrated or described herein. Furthermore, the terms "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusions, for example, processes, methods, systems, products, or devices that comprise a list of steps or units are not necessarily limited to those steps or units that are clearly listed, but can include other steps or units that are not clearly listed or inherent to such processes, methods, products, or devices.

[0041] Embodiment one

[0042] Figure 1 A flowchart of a method for constructing a user behavior prediction model is provided for the first embodiment of the application. The embodiment can be applied to the construction of a user behavior prediction model. The method can be performed by a user behavior prediction model construction device, which can be implemented in the form of hardware and / or software. The user behavior prediction model construction device can be configured in a terminal or a server with data processing functions. As shown in the figure, the method comprises the following steps. Figure 1

[0043] S110, generating an original sample set according to historical behavior data of multiple users.

[0044] In the embodiment, the historical behavior data includes behavior data of corresponding operation behaviors of multiple users on a preset platform. For example, the implementation process of a purchase operation (such as inputting a keyword, browsing search results, or clicking an item link, etc.) of a certain user when purchasing an item a, item feature information of the item a, and user feature information of the certain user, etc. Further, the preset platform can be a shopping platform with a commodity recommendation function, etc.

[0045] In the original sample, the user feature information of a certain user, the item feature information of a certain item, and the behavior set performed by the certain user on the certain item are included. There is a progressive relationship between each behavior in the behavior set.

[0046] That is, one or more user operations that meet the behavior logic are recorded in one original sample, which are performed by a certain specific user a on a certain specific item b.

[0047] ​Further, the user feature information includes at least one of the following: user age, user gender, and click data and purchase data of the setting user within a certain time; the item feature information includes at least one of the following: item category, item brand, and click volume and exposure quantity of the setting item within a certain time; and the behavior set includes at least one of the following in a progressive order: click, add-to-cart, and purchase.

[0048] In this embodiment, optionally, according to the function of the preset platform or the different prediction categories of user behaviors, the behavior set can further include the following behaviors: browsing, searching, or collecting.

[0049] In this embodiment, the progressive relationship between the behaviors can be the following: behaviors with logical relationship are sorted; for example, if the behavior set a is {click, add-to-cart, and purchase}, the behavior set a is a behavior set sorted in a progressive order; correspondingly, if the behavior set b is {add-to-cart, purchase, and click}, it is easy to understand that, under the condition of the prior art, the add-to-cart and purchase behaviors cannot occur before the click behavior, so the behavior set b is not a behavior set sorted in a progressive order.

[0050] Optionally, the original sample set is generated according to the historical behavior data of multiple users, including:

[0051] In the historical behavior data of multiple users, at least one candidate behavior set of a target user for a target item is obtained; through a logic processing module, at least one target behavior set that meets the rationality of data progression is filtered from each candidate behavior set; and at least one original sample matched with the target user is constructed according to the user feature information of the target user, the item feature information of the target item, and the at least one target behavior set.

[0052] In this embodiment, first, any user in the multiple users is selected as a target user; after the above-mentioned original sample construction behavior is performed on the current target user, any user in the multiple users except the current target user is selected as a new target user to continue to construct a new original sample, until all the users in the multiple users have constructed one or more matched original samples, or until a preset number of original samples are constructed.

[0053] In this embodiment, exemplary, if the behavior set a is {click, add to cart, purchase}, the behavior set a is a reasonable behavior set that satisfies the data progressive relationship; correspondingly, if the behavior set b is {add to cart, purchase, click}, it is easy to understand that, under the condition of the prior art, the add to cart and purchase behaviors cannot occur before the click behavior, so the behavior set b is not a reasonable behavior set that satisfies the data progressive relationship.

[0054] In this embodiment, when the candidate behavior set is screened by the logic processing module, at least one preset target behavior set that satisfies the data progressive relationship can be pre-set in the logic processing module, and then the at least one candidate behavior set is input into the logic processing module, and the logic processing module compares each candidate behavior set with the preset target behavior set in turn; if the candidate behavior set matches at least one preset target behavior set, the candidate behavior set is output as a target behavior set; if the candidate behavior set does not match all preset target behavior sets, the candidate behavior set is deleted.

[0055] Further, the user feature information matched with the user identifier, the item feature information corresponding to the item identifier and the at least one target behavior set can be spliced according to the user ID or other specific user identifier to form an original sample matched with the user identifier; further, each user in the plurality of users can construct at least one original sample matched with the user identifier; exemplary, the original sample contains information: {user a, female, item b, category c, click, add to cart, purchase}; wherein, user a in the above information is a target user, "female" is user feature information of user a, "category c" is item feature information of item b, and "click, add to cart, purchase" is a target behavior set.

[0056] S120, generating a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample respectively according to the behavior set in each original sample.

[0057] Among them, the first behavior feature matrix can be a zero matrix, and the second behavior feature matrix and the third behavior feature matrix are both non-zero matrices.

[0058] Optionally, generating a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample respectively according to the behavior set in each original sample, comprising: obtaining a current behavior set in a current processing sample in each original sample, and forming a one-dimensional behavior matrix matched with the current behavior set, wherein the number of bits L of the one-dimensional behavior matrix is fixed, and each matrix bit corresponds to a set behavior.

[0059] constructing a first basic feature matrix of L*(L-1) order, sequentially obtaining the first i-1 columns of data of the i-th column of data in the one-dimensional behavior matrix, filling the first i-1 columns of the i-th row in the first basic feature matrix, and zero-filling the remaining positions in the first basic feature matrix to obtain the first behavior feature matrix, wherein the i is initialized as 1;

[0060] constructing a second basic feature matrix of L*L order, sequentially obtaining the i-th column of data in the one-dimensional behavior matrix, filling the i-th row and the i-th column in the second basic feature matrix, and zero-filling the remaining positions in the second basic feature matrix to obtain the second behavior feature matrix;

[0061] constructing a unit matrix of L*L order as the third behavior feature matrix.

[0062] In the embodiment, first, a one-dimensional behavior matrix can be generated according to the behavior set in each original sample, and then the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix can be generated according to the one-dimensional behavior matrix.

[0063] Specifically, when the one-dimensional behavior matrix is generated according to the behavior set in each original sample, the behavior that occurs can be set as 1, and the behavior that does not occur can be set as 0 to form a one-dimensional behavior matrix composed of 0 and 1; for example, on the basis of S110, the target behavior set a satisfying the rationality of data progression relationship is {click, add to cart, purchase}, if in a behavior of a target user, the user browses the item a through the link of clicking the item, and after browsing, the user directly performs the purchase behavior without the behavior of adding to cart, then in the above behavior, the set behaviors of click and purchase are 1, and the set behavior of add to cart is 0, that is, the one-dimensional behavior matrix corresponding to the above behavior set is (1, 0, 1). It should be noted that, since in S110, at least one target behavior set satisfying the rationality of data progression relationship is filtered from each of the candidate behavior sets by the logic processing module, that is, the target behavior set is a union set having behavior rationality, further, since in a behavior, the operation of purchasing without clicking is unreasonable, in the above operation, there will be no one-dimensional behavior matrix such as (0, 0, 1) or (0, 1, 1) that does not conform to the behavior logic.

[0064] In one specific embodiment of the present embodiment, the target behavior set of the target user is set as {click, add to cart, purchase}, and the target user has click and purchase behaviors in the historical behavior data, but no add to cart behavior. After the one-dimensional behavior matrix d(1, 0, 1) formed by the above behaviors matches the current behavior set is determined, it is determined that the size of the corresponding first basic feature matrix is a 3*2 order matrix. First, the first column data of the above matrix d is obtained. Since the 0th data does not contain any data information, the first row of the first basic feature matrix has no data information. Correspondingly, the first column data of the above matrix d is obtained. The first row of the first basic feature matrix contains data 1. In this way, the third row of the first basic feature matrix contains data 1, 0. Then the above data is filled into the corresponding position of the first basic feature matrix, and the remaining position of the first basic feature matrix is zero-filled, and the first behavior feature matrix obtained is:

[0065] When the second feature matrix is constructed using the one-dimensional behavior matrix d(1, 0, 1) above, it can be determined that the size of the corresponding second basic feature matrix is a 3*3 order matrix. First, the first column data "1" in the matrix d is filled into the first row and the first column of the second basic feature matrix. Correspondingly, the second column data "0" in the matrix d is filled into the second row and the second column of the second basic feature matrix, and the third column data "1" in the matrix d is filled into the third row and the third column of the second basic feature matrix. The remaining positions of the second basic feature matrix are zero-filled, and the second behavior feature matrix is obtained as:

[0066] Further, when the third feature matrix is constructed using the one-dimensional behavior matrix d(1, 0, 1) above, a 3*3 order unit matrix can be constructed as the third behavior feature matrix.

[0067] S130, according to the user feature information of the set user in the original sample, the item feature information of the item, the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix are set to form a training sample set. The second behavior feature matrix and the third behavior feature matrix are used as labeled data in the training sample.

[0068] The training sample set includes training data and labeled data. Specifically, the training data includes user feature information of a set user, item feature information of a set item, a first behavior feature matrix, and the labeled data includes a second behavior feature matrix and a third behavior feature matrix. Further, the training data can be used to calculate a theoretical probability of user behavior prediction under a current model parameter condition. The labeled data can be used to calculate an actual probability of user behavior prediction. It should be noted that the current model parameter condition can be any condition value set randomly.

[0069] S140, training the deep recommendation model using the training sample set to obtain a user behavior prediction model.

[0070] The training sample set includes the user feature information of the set user, the item feature information of the set item, the first behavior feature matrix, the second behavior feature matrix, and the third behavior feature matrix.

[0071] Optionally, training the deep recommendation model using the training sample set to obtain a user behavior prediction model includes:

[0072] In the training sample set, a target training sample is obtained and input into the deep recommendation model. The user feature information of a target user, the item feature information of a target item, and a target first behavior feature matrix in the target training sample are processed by a sparse feature layer in the deep recommendation model to obtain an original sparse vector. The original sparse vector is processed by a dense embedding layer in the deep recommendation model to obtain a dense vector. The original sparse vector and the dense vector are subjected to logistic regression calculation by a factor decomposition layer in the deep recommendation model to obtain a behavior prediction probability matrix. A loss function is calculated according to the prediction probability matrix, a target second behavior feature matrix, and a target third behavior feature matrix in the target training sample by a loss function layer in the deep recommendation model, and the deep recommendation model is adjusted in parameters according to the loss function. The operation of obtaining the target training sample in the training sample set is returned until the user behavior prediction model is trained.

[0073] The technical scheme of the embodiment of the present application generates an original sample set according to historical behavior data of multiple users, generates a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample respectively according to a behavior set in each original sample, then forms a training sample set and labeled data according to user feature information of a set user, item feature information of a set item, the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix in the original sample, and finally trains a deep recommendation model using the training sample set to obtain a user behavior prediction model, thereby providing a new way of constructing a user behavior prediction model, and based on the new user behavior prediction model, the accuracy of user behavior prediction can be effectively improved, and the hit rate of recommended items to actual user demand can be improved, thereby effectively reducing the time consumption of users searching for interested information by themselves.

[0074] Embodiment two

[0075] Figure 2a A flowchart of a user behavior prediction method provided by the second embodiment of the present application is based on the above-mentioned embodiment and is refined in the present embodiment. In the present embodiment, training a deep recommendation model using a training sample set to obtain a user behavior prediction model is refined as follows: obtaining a target training sample in the training sample set and inputting the target training sample into the deep recommendation model; processing user feature information of a target user, item feature information of a target item and a target first behavior feature matrix in the target training sample through a sparse feature layer in the deep recommendation model to obtain an original sparse vector; processing the original sparse vector through a dense embedding layer in the deep recommendation model to obtain a dense vector; performing logistic regression calculation on the original sparse vector and the dense vector through a factor decomposition layer in the deep recommendation model to obtain a behavior prediction probability matrix; calculating a loss function according to the prediction probability matrix, a target second behavior feature matrix and a target third behavior feature matrix in the target training sample through a loss function layer in the deep recommendation model, and adjusting parameters of the deep recommendation model according to the loss function; returning to the operation of obtaining a target training sample in the training sample set until a user behavior prediction model is trained.

[0076] Correspondingly, as shown in Figure 2a the method comprises:

[0077] S210, generating an original sample set according to historical behavior data of multiple users.

[0078] The original sample includes user feature information of a set user, item feature information of a set item and a behavior set performed by the set user on the set item, and there is a progressive relationship between each behavior in the behavior set.

[0079] S220, generating a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample respectively according to the behavior set in each original sample.

[0080] S230, setting the item feature information of the item, the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix according to the user feature information of the set user in the original sample, forming a training sample set, and the second behavior feature matrix and the third behavior feature matrix are used as labeled data in the training sample.

[0081] S240, obtaining a target training sample in the training sample set and inputting the target training sample into the deep recommendation model.

[0082] As shown in Figure 2b , it should be noted that Sparse Features is the sparse feature layer in the following steps; correspondingly, Desnes Embedding is the dense embedding layer in the following steps, and HM layer is the factorization layer in the following steps; wherein, Figure 2b The Hide layer not included in the following steps is also included, and further, the Hide layer can be used for high-order cross between feature matrices and extracting deep information.

[0083] S250, processing the user feature information of the target user, the item feature information of the target item and the target first behavior feature matrix in the target training sample through the sparse feature layer in the deep recommendation model to obtain an original sparse vector.

[0084] S260, processing the original sparse vector through the dense embedding layer in the deep recommendation model to obtain a dense vector.

[0085] The dense vector can be obtained by word2vec processing of the original sparse vector in the dense embedding layer, and then the original sparse vector is converted into a dense vector.

[0086] S270, performing logistic regression calculation on the original sparse vector and the dense vector through the factorization layer in the deep recommendation model to obtain a behavior prediction probability matrix.

[0087] S280, calculating a loss function according to the prediction probability matrix, the target second behavior feature matrix and the target third behavior feature matrix in the target training sample through the loss function layer in the deep recommendation model, and adjusting the parameters of the deep recommendation model according to the loss function.

[0088] Since the current model parameter conditions can be any randomly set condition value, the predicted probability matrix of the current prediction is usually different from the actual probability of user behavior prediction calculated through labeled data. Then, the parameters are adjusted using a loss function based on the difference between the two.

[0089] S290, Return to the operation of obtaining the target training sample from the training sample set until the user behavior prediction model is trained.

[0090] like Figure 2b As shown, the input data for Sparse Features contains both sparse and numerically dense features. DesnesEmbedding can map the features to a k-dimensional vector using the FM algorithm, and perform pairwise addition of the feature vectors to obtain first-order cross weights, and pairwise dot product to obtain second-order cross weights. Then, the FM layer combines the first-order and second-order cross weights of the input data with the corresponding feature vectors, and performs data transformation on the prediction results of the FM layer and the Hide layer in the output layer to obtain the theoretical probability of user behavior prediction under the current model parameters. It should be noted that since the current model parameters can be any randomly set condition value, the predicted probability matrix of the current prediction usually differs from the actual probability of user behavior prediction calculated from the labeled data. Then, the parameters are adjusted using a loss function based on the difference between the two, and the above operation is repeated until the final difference between the two is within the allowable error range, that is, the user behavior prediction model is trained.

[0091] The technical scheme of the embodiment of the present application comprises the following steps: generating an original sample set according to historical behavior data of multiple users, generating a first behavior feature matrix, a second behavior feature matrix and a third behavior feature matrix corresponding to each original sample according to a behavior set in each original sample, forming a training sample set and labeled data according to user feature information of a set user, item feature information of a set item, the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix in the original sample, obtaining a target training sample in the training sample set, inputting the target training sample into a deep recommendation model, processing user feature information of a target user, item feature information of a target item and a target first behavior feature matrix in the target training sample through a sparse feature layer in the deep recommendation model to obtain an original sparse vector, processing the original sparse vector through a dense embedding layer in the deep recommendation model to obtain a dense vector, performing logistic regression calculation on the original sparse vector and the dense vector through a factor decomposition layer in the deep recommendation model to obtain a behavior prediction probability matrix, calculating a loss function according to the prediction probability matrix, a target second behavior feature matrix and a target third behavior feature matrix in the target training sample through a loss function layer in the deep recommendation model, adjusting parameters of the deep recommendation model according to the loss function, and finally returning to the operation of obtaining the target training sample in the training sample set until a user behavior prediction model is trained, thereby providing a new way of constructing a user behavior prediction model, based on which the accuracy of user behavior prediction can be effectively improved, and the hit rate of recommended items to actual user demand can be improved, thereby effectively reducing the time consumption of users searching for interested information by themselves.

[0092] Embodiment three

[0093] Figure 3 A flowchart of a user behavior prediction method provided by the third embodiment of the present application, the present embodiment can be applied to the case of predicting user behavior, and the method can be executed by a user behavior prediction device which can be realized in the form of hardware and / or software and can be configured in a computer or a server having a user behavior prediction function. As shown in the figure, the method comprises the following steps: Figure 3

[0094] S310, obtaining user feature information of a user to be predicted and item feature information of an item to be predicted.

[0095] ​The user feature information of the to-be-predicted user includes at least one of the following: user age, user gender, and click data and purchase data of the to-be-predicted user within a certain time period; and the item feature information of the to-be-predicted item includes at least one of the following: item category, item brand, and click volume and exposure quantity of the to-be-predicted item within a certain time period.

[0096] S320, input the user feature information of the to-be-predicted user and the item feature information of the to-be-predicted item into the user behavior prediction model trained by the method for constructing a user behavior prediction model, to obtain a behavior prediction probability matrix of the to-be-predicted user on the to-be-predicted item.

[0097] The size of the behavior prediction probability matrix is determined by the number of to-be-predicted behaviors, that is, if the to-be-predicted behaviors are n, the size of the behavior prediction probability matrix is n*n.

[0098] S330, verify whether the to-be-predicted item is recommended to the to-be-predicted user according to the behavior prediction probability matrix.

[0099] In this embodiment, it is assumed that the to-be-predicted behaviors are the probabilities of click, add-to-cart, and purchase behaviors of the to-be-predicted user on the to-be-predicted item. It is assumed that the behavior prediction probability matrix obtained by the model is According to the behavior prediction probability matrix, the probability of the click behavior is the value corresponding to the first row and the first column, that is, 0.8. Correspondingly, when the click behavior occurs, the probability of the add-to-cart behavior is the click behavior probability*the value corresponding to the second row and the second column, that is, 0.8*0.2=0.16; when the add-to-cart behavior occurs, the probability of the purchase behavior is the add-to-cart behavior probability*the value corresponding to the third row and the third column, that is, 0.16*0.3=0.048.

[0100] In this embodiment, whether the to-be-predicted item is recommended to the to-be-predicted user can be verified by setting a probability threshold, that is, when the probability of the predicted behavior reaches the threshold, the to-be-predicted item is recommended to the to-be-predicted user, otherwise, the recommendation is not performed.

[0101] In one specific embodiment of the present embodiment, on the basis of the above-mentioned behavior, when the target predicted behavior is set as clicking, it is assumed that the threshold of the clicking probability is 0.5, and since 0.8 is greater than 0.5, the to-be-predicted item can be recommended to the to-be-predicted user; further, when the target predicted behavior is set as clicking and purchasing, it is assumed that the threshold of the clicking and purchasing probability is 0.2, and the clicking and purchasing probability obtained under the current condition is 0.8*clicking weight+0.48*purchasing weight, wherein the clicking weight and the purchasing weight can be adjusted according to actual application conditions, and when the clicking and purchasing probability is greater than the preset threshold, the to-be-predicted item can be recommended to the to-be-predicted user.

[0102] The technical scheme of the present embodiment provides a new way of constructing a user behavior prediction model, and based on the new user behavior prediction model, the accuracy of user behavior prediction can be effectively improved, and the hit rate of recommended items to actual user demand can be improved, thereby effectively reducing the time consumption of users searching for interested information by themselves.

[0103] Embodiment Four

[0104] Figure 4 A structural schematic diagram of a user behavior prediction model construction device provided by the present embodiment four is shown in FIG. 4. Figure 4 As shown in the figure, the device comprises:

[0105] The original sample set generation module 410 is configured to generate an original sample set according to historical behavior data of multiple users, wherein the original sample set comprises user feature information of a set user, item feature information of a set item, and a behavior set performed by the set user on the set item, and each behavior in the behavior set has a progressive relationship.

[0106] The feature matrix generation module 420 is configured to generate a first behavior feature matrix, a second behavior feature matrix, and a third behavior feature matrix corresponding to each original sample, respectively, according to the behavior set in each original sample.

[0107] The training sample set generation module 430 is configured to form a training sample set according to the user feature information of a set user, the item feature information of a set item, the first behavior feature matrix, the second behavior feature matrix and the third behavior feature matrix in the original sample, and the second behavior feature matrix and the third behavior feature matrix are used as labeled data in the training sample.

[0108] The model training module 440 is configured to train the deep recommendation model using the training sample set to obtain a user behavior prediction model.

[0109] The technical scheme of the embodiment of the application provides a new way of constructing a user behavior prediction model, and based on the new user behavior prediction model, the accuracy of user behavior prediction can be effectively improved, and the hit rate of recommended items to actual user demand can be improved, and the time consumption of users searching for interested information by themselves can be effectively reduced.

[0110] On the basis of the above-mentioned embodiment, the original sample set generation module 410 can include:

[0111] The alternative behavior set acquisition unit is configured to acquire at least one alternative behavior set of a target user for a target item in the historical behavior data of multiple users.

[0112] The behavior set screening unit is configured to screen at least one target behavior set that meets the rationality of the data progressive relationship in each of the alternative behavior sets through a logic processing module.

[0113] The original sample construction unit is configured to construct at least one original sample matched with the target user according to the user feature information of the target user, the item feature information of the target item and the at least one target behavior set.

[0114] On the basis of the above-mentioned embodiment, the feature matrix generation module 420 can include:

[0115] The behavior matrix generation unit is configured to acquire a current behavior set in a current processing sample in each original sample and form a one-dimensional behavior matrix matched with the current behavior set, wherein the number of bits L of the one-dimensional behavior matrix is fixed, and each matrix bit corresponds to a set behavior.

[0116] The first behavior feature matrix construction unit is configured to construct a first basic feature matrix of an L*(L-1) order, sequentially obtain the first i-1 columns of data in the ith column of the one-dimensional behavior matrix, fill the first i-1 columns of the ith row in the first basic feature matrix, and fill the remaining positions in the first basic feature matrix with zeros to obtain the first behavior feature matrix, where the i is initialized as 1.

[0117] The second behavior feature matrix construction unit is configured to construct a second basic feature matrix of an L*L order, sequentially obtain the ith column of data in the one-dimensional behavior matrix, fill the ith row and the ith column in the second basic feature matrix, and fill the remaining positions in the second basic feature matrix with zeros to obtain the second behavior feature matrix.

[0118] The third behavior feature matrix construction unit is configured to construct a unit matrix of an L*L order as a third behavior feature matrix.

[0119] On the basis of the above embodiment, the model training module 440 can include:

[0120] The target training sample obtaining unit is configured to obtain a target training sample in the training sample set and input the target training sample into the deep recommendation model.

[0121] The original sparse vector generating unit is configured to process the user feature information of a target user, the item feature information of a target item, and the target first behavior feature matrix in the target training sample through a sparse feature layer in the deep recommendation model to obtain an original sparse vector.

[0122] The dense vector generating unit is configured to process the original sparse vector through a dense embedding layer in the deep recommendation model to obtain a dense vector.

[0123] The behavior prediction probability matrix obtaining unit is configured to perform logistic regression calculation on the original sparse vector and the dense vector through a factor decomposition layer in the deep recommendation model to obtain a behavior prediction probability matrix.

[0124] The parameter adjusting unit is configured to calculate a loss function according to the prediction probability matrix, a target second behavior feature matrix, and a target third behavior feature matrix in the target training sample through a loss function layer in the deep recommendation model, and adjust the parameters of the deep recommendation model according to the loss function.

[0125] The return execution unit is configured to return to execute the operation of obtaining the target training sample in the training sample set until a user behavior prediction model is trained.

[0126] The user behavior prediction model construction device provided by the embodiment of the present application can execute the user behavior prediction model construction method provided by any embodiment of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0127] Embodiment five

[0128] Figure 5 A structural schematic diagram of a user behavior prediction device provided for the embodiment five of the present application is shown in the figure. Figure 5 The device comprises:

[0129] The feature information acquisition module 510 is configured to acquire user feature information of a to-be-predicted user and item feature information of a to-be-predicted item.

[0130] The probability matrix acquisition module 520 is configured to input the user feature information of the to-be-predicted user and the item feature information of the to-be-predicted item into the user behavior prediction model trained by the user behavior prediction model construction method, and acquire a behavior prediction probability matrix of the to-be-predicted user on the to-be-predicted item.

[0131] The recommendation verification module 530 is configured to verify whether the to-be-predicted item is recommended to the to-be-predicted user according to the behavior prediction probability matrix.

[0132] The technical solution of the embodiment of the present application acquires the user feature information of the to-be-predicted user and the item feature information of the to-be-predicted item, then inputs the user feature information of the to-be-predicted user and the item feature information of the to-be-predicted item into the user behavior prediction model trained by the user behavior prediction model construction method, acquires a behavior prediction probability matrix of the to-be-predicted user on the to-be-predicted item, and finally verifies whether the to-be-predicted item is recommended to the to-be-predicted user according to the behavior prediction probability matrix, thereby providing a new way of constructing a user behavior prediction model. Based on the new user behavior prediction model, the accuracy of user behavior prediction can be effectively improved, and the hit rate of recommended items on actual user demand can be improved, thereby effectively reducing the time consumption of users searching for interested information by themselves.

[0133] Embodiment six

[0134] Figure 6A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0135] like Figure 6 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0136] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0137] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as methods for constructing user behavior prediction models, or user behavior prediction methods.

[0138] Correspondingly, the method for constructing the user behavior prediction model comprises: generating an original sample set according to historical behavior data of multiple users, the original sample comprising user feature information of a set user, item feature information of a set item, and a behavior set performed by the set user on the set item, the behaviors in the behavior set having a progressive relationship; generating a first behavior feature matrix, a second behavior feature matrix, and a third behavior feature matrix corresponding to each original sample respectively according to the behavior set in each original sample; forming a training sample set according to the user feature information of the set user, the item feature information of the set item, the first behavior feature matrix, the second behavior feature matrix, and the third behavior feature matrix in the original sample, the second behavior feature matrix and the third behavior feature matrix being used as labeled data in the training sample; training the deep recommendation model using the training sample set to obtain the user behavior prediction model.

[0139] The user behavior prediction method comprises: obtaining user feature information of a to-be-predicted user and item feature information of a to-be-predicted item; inputting the user feature information of the to-be-predicted user and the item feature information of the to-be-predicted item into the user behavior prediction model trained by the method for constructing the user behavior prediction model to obtain a behavior prediction probability matrix of the to-be-predicted user on the to-be-predicted item; and verifying whether the to-be-predicted item is recommended to the to-be-predicted user according to the behavior prediction probability matrix.

[0140] In some embodiments, the method for constructing the user behavior prediction model and the user behavior prediction method can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for data reduction in distributed training described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the method for constructing the user behavior prediction model and the user behavior prediction method by any other appropriate means, for example, by means of firmware.

[0141] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a load programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0142] Computer programs used to implement the processes of the application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer program, when executed, can cause instructions defined in the flow charts and / or block diagrams to be implemented. The computer program can be executed entirely on a machine, partially on a machine, partially on a machine as a standalone software package and partially on a remote machine or entirely on a remote machine or server.

[0143] In the context of the present application, a computer-readable storage medium can be a tangible medium that can contain or store computer programs for use by or in connection with an instruction execution system, apparatus, or device. Computer-readable storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0144] To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0145] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), blockchain network, and the Internet.

[0146] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. A server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and VPS service.

[0147] It should be understood that the various forms of flow shown above can be re-ordered, added to, or deleted from without departing from the scope of the present application. For example, the steps recited in the present application can be executed in parallel, executed in sequence, or executed in different orders, as long as the desired results of the technical solutions of the present application can be achieved, and the present application is not limited herein.

Claims

1. A method for constructing a user behavior prediction model, characterized in that, The method comprises the following steps: According to the historical behavior data of multiple users, generate a set of original samples, each original sample including user feature information of a set user, item feature information of a set item, and a set of behaviors performed by the set user on the set item, and each behavior in the set of behaviors has a progressive relationship; According to the set of behaviors in each original sample, generate a first behavior feature matrix, a second behavior feature matrix, and a third behavior feature matrix corresponding to each original sample respectively; According to the user feature information of the set user, the item feature information of the set item, the first behavior feature matrix, the second behavior feature matrix, and the third behavior feature matrix in the original sample, form a set of training samples, and the second behavior feature matrix and the third behavior feature matrix are used as labeled data in the training samples; Train the deep recommendation model using the set of training samples to obtain a user behavior prediction model; According to the set of behaviors in each original sample, generate a first behavior feature matrix, a second behavior feature matrix, and a third behavior feature matrix corresponding to each original sample respectively, which comprises: obtaining a current set of behaviors in a current sample in each original sample and forming a one-dimensional behavior matrix matching the current set of behaviors, wherein the number of bits L of the one-dimensional behavior matrix is fixed, and each matrix bit corresponds to a set behavior; construct a first basis feature matrix of L*(L-1) order, sequentially obtain the first i-1 columns of data in the i-th column of the one-dimensional behavior matrix, fill the first i-1 columns of the i-th row in the first basis feature matrix, and zero-fill the remaining positions in the first basis feature matrix to obtain the first behavior feature matrix, wherein i is initialized to 1; construct a second basis feature matrix of L*L order, sequentially obtain the i-th column of data in the one-dimensional behavior matrix, fill the i-th row and i-th column in the second basis feature matrix, and zero-fill the remaining positions in the second basis feature matrix to obtain the second behavior feature matrix; construct a unit matrix of L*L order as the third behavior feature matrix; The method comprises the following steps: In the set of training samples, obtain a target training sample and input the target training sample into the deep recommendation model; process the user feature information of a target user, the item feature information of a target item, and the target first behavior feature matrix in the target training sample through the sparse feature layer in the deep recommendation model to obtain an original sparse vector; process the original sparse vector through the dense embedding layer in the deep recommendation model to obtain a dense vector; perform logistic regression calculation on the original sparse vector and the dense vector through the factor decomposition layer in the deep recommendation model to obtain a behavior prediction probability matrix; calculate the loss function according to the prediction probability matrix, the target second behavior feature matrix, and the target third behavior feature matrix in the target training sample through the loss function layer in the deep recommendation model, and adjust the parameters of the deep recommendation model according to the loss function; return to the operation of obtaining the target training sample in the set of training samples until the user behavior prediction model is trained.

2. The method of claim 1, wherein, The user feature information includes at least one of the following: The user age, user gender, and click data and purchase data of the user within a certain time period of the user are set; The item feature information includes at least one of the following: The item category, item brand, and click volume and exposure quantity of the item within a certain time period of the item are set; The behavior set includes at least one of click, add-to-cart, and purchase in a progressive relationship.

3. The method of claim 1, wherein, According to the historical behavior data of multiple users, an original sample set is generated, including: In the historical behavior data of multiple users, at least one candidate behavior set of a target user for a target item is obtained; Through a logic processing module, at least one target behavior set that meets the rationality of the data progressive relationship is filtered from each candidate behavior set; According to the user feature information of the target user, the item feature information of the target item, and the at least one target behavior set, at least one original sample matched with the target user is constructed.

4. A user behavior prediction method characterized by, It includes: Obtaining user feature information of a user to be predicted and item feature information of an item to be predicted; Inputting the user feature information of the user to be predicted and the item feature information of the item to be predicted into a user behavior prediction model trained by the method of any one of claims 1-3 to obtain a behavior prediction probability matrix of the user to be predicted for the item to be predicted; According to the behavior prediction probability matrix, it is verified whether the item to be predicted is recommended to the user to be predicted. 5.A device for constructing a user behavior prediction model, characterized by, It includes: An original sample set generation module is configured to generate an original sample set according to historical behavior data of multiple users, and the original sample includes user feature information of a set user, item feature information of a set item, and a behavior set of the set user for the set item, and there is a progressive relationship between behaviors in the behavior set; A feature matrix generation module is configured to generate a first behavior feature matrix, a second behavior feature matrix, and a third behavior feature matrix corresponding to each original sample, respectively, according to the behavior set in each original sample; A training sample set generation module is configured to form a training sample set according to the user feature information of the set user, the item feature information of the set item, the first behavior feature matrix, the second behavior feature matrix, and the third behavior feature matrix in the original sample, and the second behavior feature matrix and the third behavior feature matrix are used as labeled data in the training sample; A model training module is configured to train a deep recommendation model using the training sample set to obtain a user behavior prediction model. The feature matrix generation module comprises: a behavior matrix generation unit, configured to obtain a current behavior set in a current processing sample from each original sample, and form a one-dimensional behavior matrix matched with the current behavior set, wherein the number of bits L of the one-dimensional behavior matrix is fixed, and each matrix bit corresponds to a set behavior; a first behavior feature matrix construction unit, configured to construct a first basic feature matrix of L*(L-1) order, sequentially obtain the first i-1 columns of data in the ith column of the one-dimensional behavior matrix, fill the first i-1 columns in the ith row of the first basic feature matrix, and perform zero padding on the remaining positions in the first basic feature matrix to obtain the first behavior feature matrix, wherein the i is initialized as 1; a second behavior feature matrix construction unit, configured to construct a second basic feature matrix of L*L order, sequentially obtain the ith column of data in the one-dimensional behavior matrix, fill the ith column in the ith row of the second basic feature matrix, and perform zero padding on the remaining positions in the second basic feature matrix to obtain the second behavior feature matrix; and a third behavior feature matrix construction unit, configured to construct a unit matrix of L*L order as the third behavior feature matrix. The model training module comprises: a target training sample acquisition unit, configured to acquire a target training sample from the training sample set, and input the target training sample into the deep recommendation model; an original sparse vector generation unit, configured to process user feature information of a target user, item feature information of a target item and a target first behavior feature matrix in the target training sample through a sparse feature layer in the deep recommendation model to obtain an original sparse vector; a dense vector generation unit, configured to process the original sparse vector through a dense embedding layer in the deep recommendation model to obtain a dense vector; a behavior prediction probability matrix acquisition unit, configured to perform logistic regression calculation on the original sparse vector and the dense vector through a factor decomposition layer in the deep recommendation model to obtain a behavior prediction probability matrix; a parameter adjustment unit, configured to calculate a loss function according to the prediction probability matrix, a target second behavior feature matrix and a target third behavior feature matrix in the target training sample through a loss function layer in the deep recommendation model, and adjust parameters of the deep recommendation model according to the loss function; and a return execution unit, configured to return to execute the operation of acquiring the target training sample from the training sample set until a user behavior prediction model is trained.

6. A user behavior prediction apparatus characterized by comprising: The method comprises: obtaining user feature information of a to-be-predicted user and item feature information of a to-be-predicted item; inputting the user feature information of the to-be-predicted user and the item feature information of the to-be-predicted item into a user behavior prediction model trained by the method in any one of claims 1-3 to obtain a behavior prediction probability matrix of the to-be-predicted user on the to-be-predicted item; verifying whether to recommend the to-be-predicted item to the to-be-predicted user according to the behavior prediction probability matrix.

7. An electronic device, comprising: The electronic device comprises: at least one processor; and A memory connected in communication with the at least one processor; wherein The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the method for constructing a user behavior prediction model according to any one of claims 1-3, or perform the user behavior prediction method according to claim 4.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to perform the method for constructing a user behavior prediction model according to any one of claims 1-3, or perform the user behavior prediction method according to claim 4 when executed.

Images