Article data processing method and device, electronic equipment and storage medium

Abstract

The invention discloses an article data processing method and device, electronic equipment and a storage medium, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of obtaining a corresponding article set and a target user group in response to an article processing instruction to determine attribute parameters of each article in the article set, and querying attribute parameters of each user in the target user group; based on the attribute parameter of each article and the attribute parameter of each user, calculating the score of each article by each user; the scores are weighted through all the weight coefficients in the weight coefficient set, the weighted scores are fused, and the scoring result of each article is obtained; and predicting a corresponding conversion probability based on the scoring result of each article, so as to generate recommendation information of the article set based on the conversion probability. According to the embodiment, the problem that due to the fact that the private domain user group changes frequently, accurate portraying of the user group is difficult, and the accuracy of recommended items of the private domain user group is low can be solved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, electronic device and storage medium for processing item data. Background Art

[0002] In the field of e-commerce operations, to better serve users, business personnel can direct users from the public domain to the private domain of individuals or enterprises to provide personalized services to the user groups in the private domain. For example, to facilitate users' selection from a large number of items, item recommendations can be made for private domain user groups. In related technologies, item recommendations for user groups in private domains are usually selected based on the aggregated profile of the user group, and then a list of items is generated and pushed to the user group. However, due to the frequent changes in private domain user groups, it is difficult to accurately profile the user group, resulting in a low accuracy rate of recommended items for private domain user groups. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a method, device, electronic device and storage medium for item data processing, which can solve the problem that the accuracy of recommended items for private domain user groups is low due to the frequent changes in private domain user groups and the difficulty in accurately profiling the user groups.

[0004] To achieve the above objective, according to one aspect of an embodiment of the present invention, a method for processing item data is provided.

[0005] A method for processing item data according to an embodiment of the present invention includes: in response to an item processing instruction, obtaining a corresponding item set and a target user group to determine attribute parameters of each item in the item set, and querying attribute parameters of each user in the target user group;

[0006] Calculating a rating of each user for each item based on the attribute parameters of each item and the attribute parameters of each user;

[0007] Invoking a preset calculation model to weight the scores using each weight coefficient in the weight coefficient set, and fusing the weighted scores to obtain a score result for each of the items;

[0008] A preset prediction model is called to predict the corresponding conversion probability based on the scoring result of each of the items, so as to generate recommendation information for the item set based on the conversion probability.

[0009] In one embodiment, calculating the rating of each user for each item based on the attribute parameters of each item and the attribute parameters of each user includes:

[0010] Generate a corresponding item vector based on the attribute parameters of each item, and generate a user vector based on the attribute parameters of each user;

[0011] The similarity between the item vector of each item and the user vector of each user is calculated to determine a rating vector for each item by each user.

[0012] In yet another embodiment, the weighted scores are combined to obtain a score result for each item, including:

[0013] The weighted scores of the weight coefficients are concatenated, and the preset linear layer is called to linearize the concatenated scores to obtain the score result of each item.

[0014] In yet another embodiment, before weighting the score using each weight coefficient in the weight coefficient set, the method further includes:

[0015] An activity information set corresponding to the target user group of each user is obtained to generate at least one activity weight coefficient, and the activity weight coefficient set is updated.

[0016] In yet another embodiment, obtaining an activity information set of each user corresponding to the target user group to generate at least one activity weight coefficient includes:

[0017] Extracting, from the active information set, first attribute information of each user logging into the target user group and second attribute information of each user operating recommended information in the target user group;

[0018] A preset generation model is called to generate a first activity weight coefficient based on the first attribute information and a second activity weight coefficient based on the second attribute information.

[0019] In yet another embodiment, generating recommendation information for the item set based on the conversion probability includes:

[0020] Determining an arrangement order of items in the item set based on the conversion probability;

[0021] Obtain parameter values ​​of list parameters corresponding to each of the items to generate item recommendation list information according to the arrangement order.

[0022] In yet another embodiment, before responding to the item handling instruction, the method further comprises:

[0023] Obtaining basic information and user behavior information of each user in each user group to extract user characteristics of each user;

[0024] Obtaining basic information and operation information of each item associated with each user group to extract item features of each item;

[0025] Obtaining a first mapping relationship between successfully converted items and corresponding users, and a second mapping relationship between each of the user groups and the users included therein;

[0026] Constructing model training data based on the user characteristics, the item characteristics, the first mapping relationship, and the second mapping relationship;

[0027] A preset training procedure is performed on a preset model based on the training data, wherein the preset model includes the calculation model and the prediction model.

[0028] To achieve the above objective, according to another aspect of an embodiment of the present invention, a device for processing item data is provided.

[0029] An apparatus for processing item data according to an embodiment of the present invention includes: an acquisition unit configured to, in response to an item processing instruction, acquire a corresponding item set and a target user group, thereby determining attribute parameters of each item in the item set and querying attribute parameters of each user in the target user group;

[0030] a calculation unit, configured to calculate a score of each user for each of the items based on the attribute parameters of each of the items and the attribute parameters of each of the users;

[0031] The calculation unit is further configured to call a preset calculation model to weight the scores using each weight coefficient in the weight coefficient set, and to combine the weighted scores to obtain a score result for each item;

[0032] The generating unit is configured to call a preset prediction model to predict a corresponding conversion probability based on the scoring result of each of the items, so as to generate recommendation information for the item set based on the conversion probability.

[0033] In one embodiment, the computing unit is specifically configured to:

[0034] Generate a corresponding item vector based on the attribute parameters of each item, and generate a user vector based on the attribute parameters of each user;

[0035] The similarity between the item vector of each item and the user vector of each user is calculated to determine a score for each item given by each user.

[0036] In yet another embodiment, the computing unit is specifically configured to:

[0037] The weighted scores of the weight coefficients are concatenated, and the preset linear layer is called to linearize the concatenated scores to obtain the score result of each item.

[0038] In yet another embodiment, the generating unit is further configured to:

[0039] An activity information set corresponding to the target user group of each user is obtained to generate at least one activity weight coefficient, and the activity weight coefficient set is updated.

[0040] In yet another embodiment, the generating unit is specifically configured to:

[0041] Extracting, from the active information set, first attribute information of each user logging into the target user group and second attribute information of each user operating recommended information in the target user group;

[0042] A preset generation model is called to generate a first activity weight coefficient based on the first attribute information and a second activity weight coefficient based on the second attribute information.

[0043] In yet another embodiment, the generating unit is specifically configured to:

[0044] Determining an arrangement order of items in the item set based on the conversion probability;

[0045] Obtain parameter values ​​of list parameters corresponding to each of the items to generate item recommendation list information according to the arrangement order.

[0046] In yet another embodiment, the apparatus further comprises:

[0047] An extraction unit is configured to obtain basic information and user behavior information of each user in each user group to extract user characteristics of each user; and obtain basic information and operation information of each item associated with each user group to extract item characteristics of each item;

[0048] A construction unit is configured to obtain a first mapping relationship between successfully converted items and corresponding users, and a second mapping relationship between each user group and the users included therein; and construct model training data based on the user characteristics, the item characteristics, the first mapping relationship, and the second mapping relationship;

[0049] A training unit is used to execute a preset training program on a preset model based on the training data, wherein the preset model includes the calculation model and the prediction model.

[0050] To achieve the above objective, according to another aspect of an embodiment of the present invention, an electronic device is provided.

[0051] An electronic device according to an embodiment of the present invention includes: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for processing item data provided by the embodiment of the present invention.

[0052] To achieve the above objective, according to another aspect of an embodiment of the present invention, a computer-readable medium is provided.

[0053] A computer-readable medium according to an embodiment of the present invention stores a computer program, which, when executed by a processor, implements the method for processing item data provided by an embodiment of the present invention.

[0054] To achieve the above objective, according to another aspect of the embodiments of the present invention, a computer program product is provided.

[0055] A computer program product according to an embodiment of the present invention includes a computer program, which, when executed by a processor, implements the method for processing item data provided by an embodiment of the present invention.

[0056] One embodiment of the above invention has the following advantages or beneficial effects: In this embodiment of the present invention, for a target user group and item set, the user's rating of the item can be first calculated based on various attribute parameters, that is, the association between the user and the item. The rating can then be weighted using multi-dimensional weight coefficients, that is, the degree of influence of the user on the item rating can be adjusted from multiple different dimensions. After the weighted ratings are integrated, the conversion probability of the item is calculated using a prediction model, thereby generating recommendation information for the item set. In this way, in this embodiment of the present invention, after calculating the association between the user attribute parameters and the item attributes, the degree of influence of the user on the item rating is also adjusted using multi-dimensional weight coefficients to calculate the conversion probability of the item in the target user group. This allows item recommendations to be made without relying on user profiles, thereby improving the accuracy of item recommendations.

[0057] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings are provided for a better understanding of the present invention and are not intended to limit the present invention.

[0059] Figure 1 is a schematic diagram of a main process of a method for processing item data according to an embodiment of the present invention;

[0060] Figure 2 is a schematic diagram of another main process of the method for processing item data according to an embodiment of the present invention;

[0061] Figure 3 is a schematic diagram of main units of an apparatus for processing item data according to an embodiment of the present invention;

[0062] Figure 4 is an exemplary system architecture diagram in which embodiments of the present invention may be applied;

[0063] Figure 5 It is a schematic diagram of the structure of a computer system suitable for implementing the embodiment of the present invention. DETAILED DESCRIPTION

[0064] The following description of exemplary embodiments of the present invention is made in conjunction with the accompanying drawings, in which various details of the embodiments of the present invention are included to facilitate understanding. These details should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0065] It should be noted that the embodiments and features of the embodiments of the present invention can be combined with each other without conflict. The acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations.

[0066] An embodiment of the present invention provides an item data processing system, which can be used in a scenario of recommending items to a user group, and specifically can be used in a scenario of recommending items that are remembered by users in a private domain user group.

[0067] In the e-commerce sector, operations management personnel can connect with users in the public domain through existing instant messaging tools, such as IM tools. They can communicate with users through IM tools, public platforms, or one-on-one messaging, thereby directing public users to the private domain. To facilitate item recommendations for private users, user groups can be established for communication with private users, such as IM group chats, enabling efficient item recommendations. However, user groups frequently change, and user preferences vary. Therefore, item recommendations must be highly accurate to maximize item conversion probability. Item conversion probability refers to the probability that a recommended item will be converted into an item ordered by the user.

[0068] The embodiment of the present invention provides a method for processing item data, which can be executed by an item data processing system. Figure 1 As shown, the method includes:

[0069] S101: In response to an item processing instruction, a corresponding item set and a target user group are obtained to determine attribute parameters of each item in the item set and query attribute parameters of each user in the target user group.

[0070] The item processing instruction can be automatically triggered by the system or sent by another system. For example, the item data processing system can set an item recommendation time so that the item processing instruction is automatically triggered when the item recommendation time arrives; or the operation management system can send the item processing instruction to the item data processing system so that the item data processing system can execute the embodiments of the present invention.

[0071] An item processing instruction can include an item set, representing the set of items eligible for recommendation. Items within the item set can be determined using various methods, such as sales volume, listing time, or discount periods. The item set can also include identification information for each item to identify the specific item. After obtaining the item set, this step can retrieve the attribute parameters of each item in the item set based on the item identification.

[0072] The attribute parameters of an item can be set based on demand, and may include, for example, price, category, brand, click records, conversion records, etc.

[0073] The target user group represents the user group for which the item is recommended. It can be included in the item processing instruction or pre-configured by the item data processing system. The target user group can include the user ID of the user, so that the attribute parameters of each user in the target user group can be queried based on the user ID.

[0074] The user's attribute parameters can be set based on demand, for example, they can include the user's browsing history, purchase history, viewing history, payment history, return history, etc.

[0075] It should be noted that, in the embodiment of the present invention, there may be multiple user groups for which item recommendations are required. In this case, each user group may be determined as a target user group in turn, and the embodiment of the present invention may be executed in turn to determine item recommendation information.

[0076] S102: Calculate each user's rating of each item based on the attribute parameters of each item and the attribute parameters of each user.

[0077] The user's rating of each item can represent the association between the user and the item, that is, the possibility of the item being converted into an item ordered by the user, thereby determining the initial association between the user and the item.

[0078] Specifically, in an embodiment of the present invention, the user's rating of an item can be represented by the similarity of the attribute parameters between the user and the item. Therefore, this step can be specifically performed as follows: generating a corresponding item vector based on the attribute parameters of each item, and generating a user vector based on the attribute parameters of each user; calculating the similarity between the item vector of each item and the user vector of each user to determine the rating of each user for each item.

[0079] In this step, the embedding model can be used to generate item vectors and user vectors. The item vector can represent the feature vector of the item, and the user vector can represent the feature vector of the user. For ease of calculation, in this embodiment of the present invention, the number of dimensions of the item attribute parameters and the number of dimensions of the user attribute parameters are the same, that is, the item vector and the user vector have the same dimensions.

[0080] In the embodiment of the present invention, the method for calculating the similarity between the item vector of each item and the user vector of each user is not limited. For example, the similarity can be calculated through dot product calculation, cosine similarity calculation, factorization machine, logistic regression, etc.

[0081] S103: calling a preset calculation model to weight the scores respectively using each weight coefficient in the weight coefficient set, and fusing the weighted scores to obtain a score result for each item.

[0082] The preset calculation model can be pre-trained, and the network architecture of the preset calculation model can be configured based on demand. In this step, the weight coefficient set can include multi-dimensional weight coefficients to adjust the influence of each user on the item conversion probability from multiple dimensions. Weighting the score using each weight coefficient in the weight coefficient set can be specifically performed by multiplying the score obtained in step S101 by each weight coefficient to obtain a weighted score. The multiple dimensions can be configured based on demand, such as user activity, user type, etc.

[0083] In one embodiment of the present invention, since users with different levels of activity have varying levels of demand for items, weight coefficients can be determined based on the user's activity level within the user group. Specifically, the user's influence on the item conversion probability is adjusted based on their activity level, causing inactive users to have their item ratings downgraded and active users to have their ratings upgraded. Therefore, prior to this step, a set of activity information for each user's corresponding target user group can be obtained to generate at least one activity weight coefficient, which is then updated into the weight coefficient set.

[0084] Specifically, the active information set can include a variety of information representing user activity, such as information about users logging into the target user group, information about users browsing messages in the target user group, information about users manipulating recommended information in the target user group, and so on. Therefore, in this step, at least one weight coefficient can be generated based on different dimensions. Therefore, the step of generating at least one active weight coefficient can be specifically performed by: extracting first attribute information of each user logging into the target user group and second attribute information of each user manipulating recommended information in the target user group from the active information; calling a preset generation model to generate a first active weight coefficient based on the first attribute information and a second active weight coefficient based on the second attribute information. The preset generation model can be set based on demand. For example, for the first attribute information of users logging into the target user group, the frequency and duration of users' logins over a period of time can be counted, and then normalized and summed to serve as the first active weight coefficient. For the second attribute information of users manipulating recommended information in the target user group, the number of times users viewed and ordered recommended items over a period of time can be counted, and then normalized and summed to serve as the second active weight coefficient.

[0085] In the embodiment of the present invention, for each item, the weighted ratings of each user in the target user group can be obtained, and then the weighted ratings can be fused to obtain the rating result of the item.

[0086] Specifically, the weighted score fusion in this step can be specifically performed by concatenating the scores weighted by each weight coefficient, and linearizing the concatenated scores using a preset linear layer to obtain the score result for each item. The concatenation method is not limited.

[0087] In another embodiment, the embodiment of the present invention can also be implemented by a multi-head attention mechanism model. The multi-head attention mechanism model is pre-trained, which may include W q 、W k and W v Weight coefficient, in the multi-head attention mechanism model, first calculate the vector value of the three dimensions q, k, and v for the input, that is, the weighted value of these three dimensions, so these three dimensions correspond to the weight coefficient W q 、W k and W v Where q is the query vector in the attention mechanism, representing the query "word" vector, k is the key vector in the attention mechanism, representing the key vector in the vocabulary, and v is the value vector in the attention mechanism, representing the value vector in the vocabulary. The specific calculation method can be shown in Formula 1, where s represents the input vector.

[0088] q=s*W q ,k=s*Wk ,v=s*W v (1)

[0089] In the multi-head attention mechanism model, after calculating q, k, and v, they can be input into a preset concat layer for splicing, and then input into a preset linear layer to output the calculation result. Therefore, in this embodiment of the present invention, the score calculated in step S102 can be input into the multi-head attention mechanism model to calculate the score result for each item.

[0090] S104: calling a preset prediction model to predict the corresponding conversion probability based on the rating result of each item, so as to generate recommendation information of the item set based on the conversion probability.

[0091] The prediction model can be pre-set and implemented using a multi-layer perceptron. The multi-layer perceptron, consisting of multiple linear layers and activation functions, can convert the scoring output by the attention mechanism into an item conversion probability, that is, an estimated probability that the item will be sold by users in the target user group.

[0092] Specifically, the prediction model may include an input layer, multiple computational layers, and an output layer. The computational layer may be composed of multiple nodes (neurons). The scoring result for each item is input into the prediction model through the output layer, then sequentially calculated by each computational layer, and finally output by the output layer. Each computational layer may first perform a linear calculation using a weighted summation on the input data from the previous layer, and then perform a nonlinear transformation using a preset activation function. After the final computational layer completes the calculation, it outputs the result, i.e., the conversion probability corresponding to each item. The prediction model may be pre-trained, with the computational parameters of each computational layer determined through model training.

[0093] After determining the conversion probability of each item, items with high conversion probabilities can be prioritized for recommendation. For private domain user groups, item recommendations are typically made in the form of item lists. Therefore, in this step, the order of items in the item set can be determined based on the conversion probability. The parameter values ​​for the list parameters required to generate the list for each item can then be obtained. This allows for the generation of recommended item lists based on the order of their order, i.e., recommendation information for the item set.

[0094] It should be noted that, in the embodiment of the present invention, the item data processing system can push the generated recommendation information to the target user group immediately via a preset interface by means of message push or the like, so that users in the target user group can view or place an order.

[0095] In an embodiment of the present invention, after calculating the correlation between each user attribute parameter and item attribute, the influence of the user on the item rating is adjusted by multi-dimensional weight coefficients to calculate the conversion probability of the item in the target user group. This allows item recommendations to be made without relying solely on user profiles, thereby improving the accuracy of item recommendations.

[0096] It should be noted that in the embodiment of the present invention, when using the multi-head attention mechanism model, the multi-head attention mechanism model needs to be trained in advance to obtain W q 、W k and W v Parameter values. Specifically, training samples can be constructed through the group relationships in the private domain user group and the transaction records of the users in the group. Since each private domain user group and the users in the group can be associated through identifiers, the group relationship can be determined through the association relationship, and then the attribute parameters of each user and recommended item can be obtained as training data. The training data can be split into training data and test data, specifically in an 8:2 ratio. A certain proportion of new data (that is, data that has not appeared in both training and testing) can also be mixed into the test data to test the generalization performance of the model.

[0097] In an embodiment of the present invention, the loss function of the model training can be mainly based on the negative log-likelihood function, and the L1 and L2 regularization terms can be appropriately added to suppress the overfitting degree of the model, which can be specifically shown in Formula 2.

[0098]

[0099] In formula 2, y i represents the actual conversion probability of the i-th item (i∈[1,N]) (0=not converted, 1=converted), is an estimate of the conversion probability of the i-th item, λ1 and λ2 represent the L1 and L2 regularization coefficients respectively (λ1,λ1∈[0,+∞)). The larger the two coefficients are, the greater the weight of the regularization term is, and the less likely the model is to overfit. i Represents all weight coefficients of the i-th item.

[0100] During training, model optimizers can be selected that utilize gradient descent as their core and dynamically optimize the learning rate based on momentum, such as adagrad and adam. During model training, it is necessary to appropriately explore the impact of the initial learning rate, epochs (the number of times the dataset is fully traversed during training), and the size of the batch training data on the decline of the model loss function. The best-performing model instance should be selected based on the model's performance on the area under the curve (AUC). The AUC metric is a commonly used metric in search and recommendation ranking scenarios. A higher AUC indicates that positive samples (transformed data) are more likely to be ranked higher than negative samples (untransformed data) in the model's prediction score. When predicting models, the output of the model through the activation function can be directly used to sort the list.

[0101] It should be noted that, in the embodiment of the present invention, the calculation model and the prediction model can also be combined as a whole for model training, that is, the calculation model output is used as the prediction model input to train the quantity model combination.

[0102] Specifically, model training can be performed as follows: obtaining basic information and user behavior information of each user in each user group to extract user characteristics of each user; obtaining basic information and operated information of each item associated with each user group to extract item characteristics of each item; obtaining a first mapping relationship between successfully converted items and corresponding users, and a second mapping relationship between each user group and the users included therein; constructing model training data based on user characteristics, item characteristics, the first mapping relationship and the second mapping relationship; and executing a preset training program on the preset model based on the training data.

[0103] The preset model is a combination of pre-built calculation models and prediction models. The user group represents the private domain user group, and the basic information and user behavior information of the users in each user group after desensitization can be obtained. The basic information may include gender, age, city, etc. The user behavior information may include application behavior (such as browsing items, category records, adding to cart records, client type, etc.), ordering behavior (such as order category, order timestamp, order item price, order item quantity, etc.), payment behavior, return and after-sales behavior (such as return frequency, return items, category labels, etc.). The item identification associated with each user group is the recommended item in the user group. The basic information of the item may include the attributes of the item, such as price, category, brand, label, etc. The operation information of the item may include click records (number of clicks, average stay time, etc.), conversion records (such as historical click-through rate), and after-sales (historical return and exchange rate, number of reviews, praise rate, etc.). At the same time, based on the transaction records of items in the private domain, successfully converted items can also be obtained to establish a first mapping relationship based on the users who successfully converted the items. In addition, based on the users included in each user group, a second mapping relationship can be established between each user group and the users included. Combining the first mapping relationship and the second mapping relationship, a relationship table between user group-user-successfully converted items can be established, and then the training samples of the preset model can be constructed by combining the characteristics of the items and the characteristics of the users. The training program of the preset model can be pre-set, specifically, the training method of the multi-head attention mechanism model mentioned above can be combined with the training method of the neural network model.

[0104] The following combination Figure 2 The system architecture shown in FIG. 1 is used to specifically describe the method for processing item data in an embodiment of the present invention. Figure 2 As shown, the method includes:

[0105] S201: In response to an item processing instruction, a corresponding item set and a target user group are obtained to determine attribute parameters of each item in the item set and query attribute parameters of each user in the target user group.

[0106] S202: Generate a corresponding item vector based on the attribute parameters of each item, and generate a user vector based on the attribute parameters of each user.

[0107] S203: Calculate the similarity between the item vector of each item and the user vector of each user to determine a rating vector for each user for each item.

[0108] S204: Calling a preset multi-attention mechanism model to calculate the rating result of each item based on the rating vector of each item.

[0109] S205: Calling a preset prediction model to predict the corresponding conversion probability based on the rating result of each item.

[0110] S206: Determine the arrangement order of each item in the item set based on the conversion probability.

[0111] S207: Obtain parameter values ​​of list parameters corresponding to each item to generate item recommendation list information in order of arrangement.

[0112] It should be noted that the data processing principle in the embodiment of the present invention is the same as Figure 1 The corresponding data processing principles in the illustrated embodiments are the same and will not be described again here.

[0113] In order to solve the problems existing in the prior art, the embodiment of the present invention provides an apparatus 300 for processing item data, such as Figure 3 As shown, the device 300 includes:

[0114] An acquisition unit 301 is configured to acquire a corresponding item set and a target user group in response to an item processing instruction, to determine attribute parameters of each item in the item set, and to query attribute parameters of each user in the target user group;

[0115] A calculation unit 302 is configured to calculate a rating of each user for each item based on the attribute parameters of each item and the attribute parameters of each user;

[0116] The calculation unit 302 is further configured to call a preset calculation model to weight the scores using each weight coefficient in the weight coefficient set, and to combine the weighted scores to obtain a score result for each item.

[0117] The generating unit 303 is configured to call a preset prediction model to predict a corresponding conversion probability based on the scoring result of each of the items, so as to generate recommendation information for the item set based on the conversion probability.

[0118] It should be understood that the manner in which the embodiments of the present invention are implemented is different from the manner in which the embodiments of the present invention are implemented. Figure 1 The methods of the illustrated embodiments are the same and will not be described again here.

[0119] In one embodiment, the computing unit 302 is specifically configured to:

[0120] Generate a corresponding item vector based on the attribute parameters of each item, and generate a user vector based on the attribute parameters of each user;

[0121] The similarity between the item vector of each item and the user vector of each user is calculated to determine a score for each item given by each user.

[0122] In yet another embodiment, the computing unit 302 is specifically configured to:

[0123] The weighted scores of the weight coefficients are concatenated, and the preset linear layer is called to linearize the concatenated scores to obtain the score result of each item.

[0124] In yet another embodiment, the generating unit 303 is further configured to:

[0125] An activity information set corresponding to the target user group of each user is obtained to generate at least one activity weight coefficient, and the activity weight coefficient set is updated.

[0126] In yet another embodiment, the generating unit 303 is specifically configured to:

[0127] Extracting, from the active information set, first attribute information of each user logging into the target user group and second attribute information of each user operating recommended information in the target user group;

[0128] A preset generation model is called to generate a first activity weight coefficient based on the first attribute information and a second activity weight coefficient based on the second attribute information.

[0129] In yet another embodiment, the generating unit 303 is specifically configured to:

[0130] Determining an arrangement order of items in the item set based on the conversion probability;

[0131] Obtain parameter values ​​of list parameters corresponding to each of the items to generate item recommendation list information according to the arrangement order.

[0132] In yet another embodiment, the apparatus 300 further includes:

[0133] An extraction unit is configured to obtain basic information and user behavior information of each user in each user group to extract user characteristics of each user; and obtain basic information and operation information of each item associated with each user group to extract item characteristics of each item;

[0134] A construction unit is configured to obtain a first mapping relationship between successfully converted items and corresponding users, and a second mapping relationship between each user group and the users included therein; and construct model training data based on the user characteristics, the item characteristics, the first mapping relationship, and the second mapping relationship;

[0135] A training unit is used to execute a preset training program on a preset model based on the training data, wherein the preset model includes the calculation model and the prediction model.

[0136] It should be understood that the manner in which the embodiments of the present invention are implemented is different from the manner in which the embodiments of the present invention are implemented. Figure 1 、 2The methods of the illustrated embodiments are the same and will not be described again here.

[0137] In an embodiment of the present invention, after calculating the correlation between each user attribute parameter and item attribute, the influence of the user on the item rating is adjusted by multi-dimensional weight coefficients to calculate the conversion probability of the item in the target user group. This allows item recommendations to be made without relying solely on user profiles, thereby improving the accuracy of item recommendations.

[0138] According to an embodiment of the present invention, an electronic device and a readable storage medium are further provided.

[0139] An electronic device according to an embodiment of the present invention includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor so that the at least one processor executes the method for processing item data provided by the embodiment of the present invention.

[0140] Figure 4 An exemplary system architecture 400 is shown to which the method or apparatus for item data processing according to an embodiment of the present invention may be applied.

[0141] like Figure 4 As shown, system architecture 400 may include terminal devices 401, 402, 403, a network 404, and a server 405. Network 404 is used to provide a medium for communication links between terminal devices 401, 402, 403 and server 405. Network 404 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0142] Users can use terminal devices 401, 402, 403 to interact with server 405 via network 404 to receive or send messages, etc. Various client applications can be installed on terminal devices 401, 402, 403.

[0143] The terminal devices 401 , 402 , and 403 may be, but are not limited to, smart phones, tablet computers, laptop computers, and desktop computers, etc.

[0144] The server 405 may be a server that provides various services. The server may analyze and process received data such as product information query requests, and feed back processing results (such as product information—only an example) to the terminal device.

[0145] It should be noted that the method for processing item data provided in the embodiment of the present invention is generally executed by the server 405 , and accordingly, the device for processing item data is generally set in the server 405 .

[0146] It should be understood that Figure 4 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0147] Reference below Figure 5 , which shows a schematic structural diagram of a computer system 500 suitable for implementing an embodiment of the present invention. Figure 5 The computer system shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0148] like Figure 5 As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage unit 508 into a random access memory (RAM) 503. Various programs and data required for the operation of the system 500 are also stored in the RAM 503. The CPU 501, ROM 502, and RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0149] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, and the like; an output section 507 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 508 including a hard disk; and a communication section 509 including a network interface card such as a LAN card or a modem. The communication section 509 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 510 as needed, so that computer programs read therefrom can be installed into the storage section 508 as needed.

[0150] In particular, according to the embodiments disclosed in the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 509, and / or installed from a removable medium 511. When the computer program is executed by the central processing unit (CPU) 501, the above-mentioned functions defined in the system of the present invention are performed.

[0151] It should be noted that the computer-readable medium described in the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media can include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated data signal can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical fiber cable, RF, or any suitable combination thereof.

[0152] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a unit, program segment, or a part of code, and the above-mentioned unit, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0153] The units described in the embodiments of the present invention may be implemented in software or hardware. The units described may also be provided in a processor. For example, a processor may be described as comprising an acquisition unit, a calculation unit, and a generation unit. The names of these units do not, in some cases, limit the units themselves. For example, an acquisition unit may also be described as a "unit for acquiring a function."

[0154] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiments, or may exist independently and not incorporated into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to perform the method for processing item data provided by the present invention.

[0155] As another aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method for processing item data provided in an embodiment of the present invention.

[0156] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for processing item data, characterized in that: include: In response to the item processing instruction, obtaining a corresponding item set and a target user group to determine attribute parameters of each item in the item set and query attribute parameters of each user in the target user group; Calculating a rating of each user for each item based on the attribute parameters of each item and the attribute parameters of each user; Invoking a preset calculation model to weight the scores using each weight coefficient in the weight coefficient set, and fusing the weighted scores to obtain a score result for each of the items; A preset prediction model is called to predict the corresponding conversion probability based on the scoring result of each of the items, so as to generate recommendation information for the item set based on the conversion probability.

2. The method according to claim 1, characterized in that Calculating a score of each user for each item based on the attribute parameters of each item and the attribute parameters of each user, including: Generate a corresponding item vector based on the attribute parameters of each item, and generate a user vector based on the attribute parameters of each user; The similarity between the item vector of each item and the user vector of each user is calculated to determine a rating vector for each item by each user.

3. The method according to claim 1, characterized in that The weighted scores are combined to obtain the score results for each item, including: The weighted scores of the weight coefficients are concatenated, and the preset linear layer is called to linearize the concatenated scores to obtain the score result of each item.

4. The method according to claim 1, wherein Before weighting the score by each weight coefficient in the weight coefficient set, the method further includes: An activity information set corresponding to the target user group of each user is obtained to generate at least one activity weight coefficient, and the activity weight coefficient set is updated.

5. The method according to claim 4, characterized in that Acquiring an activity information set of each user corresponding to the target user group to generate at least one activity weight coefficient includes: Extracting, from the active information set, first attribute information of each user logging into the target user group and second attribute information of each user operating recommended information in the target user group; A preset generation model is called to generate a first activity weight coefficient based on the first attribute information and a second activity weight coefficient based on the second attribute information.

6. The method according to claim 1, characterized in that Generating recommendation information for the item set based on the conversion probability includes: Determining an arrangement order of items in the item set based on the conversion probability; Obtain parameter values ​​of list parameters corresponding to each of the items to generate item recommendation list information according to the arrangement order.

7. The method according to claim 1, characterized in that Before responding to the item handling instruction, the method further includes: Obtaining basic information and user behavior information of each user in each user group to extract user characteristics of each user; Obtaining basic information and operation information of each item associated with each user group to extract item features of each item; Obtaining a first mapping relationship between successfully converted items and corresponding users, and a second mapping relationship between each of the user groups and the users included therein; Constructing model training data based on the user characteristics, the item characteristics, the first mapping relationship, and the second mapping relationship; A preset training procedure is performed on a preset model based on the training data, wherein the preset model includes the calculation model and the prediction model.

8. An apparatus for processing item data, characterized in that: include: an acquisition unit, configured to acquire a corresponding item set and a target user group in response to an item processing instruction, to determine attribute parameters of each item in the item set, and to query attribute parameters of each user in the target user group; a calculation unit, configured to calculate a score of each user for each of the items based on the attribute parameters of each of the items and the attribute parameters of each of the users; The calculation unit is further configured to call a preset calculation model to weight the scores using each weight coefficient in the weight coefficient set, and to combine the weighted scores to obtain a score result for each item; The generating unit is configured to call a preset prediction model to predict a corresponding conversion probability based on the scoring result of each of the items, so as to generate recommendation information for the item set based on the conversion probability.

9. An electronic device, characterized in that: include: one or more processors; a storage device 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 method according to any one of claims 1 to 7.

10. A computer-readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images