Product recommendation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a product recommendation method and device, electronic equipment and a storage medium, belongs to the technical field of artificial intelligence, and is suitable for financial science and technology scenes and medical insurance scenes. The method comprises the steps of obtaining product recommendation sample data of candidate products; performing model training on at least two preset original product recommendation dichotomy models based on the product recommendation sample data to obtain at least two target product recommendation dichotomy models and corresponding model confusion matrixes; performing model weight distribution based on the model confusion matrix to obtain a target model weight matrix; performing product prediction on the user information of the target object through a plurality of target product recommendation dichotomy models to obtain an original product recommendation probability; and carrying out product screening based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence, and carrying out product recommendation on a target object. According to the embodiment of the invention, the comprehensiveness of product recommendation can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology and is applicable to financial technology scenarios and medical insurance scenarios, and in particular to a product recommendation method and device, electronic device, and storage medium. Background Art

[0002] Product recommendation is an AI technology that can filter and recommend products that customers may be interested in based on their user and product information. This technology can be applied across multiple scenarios. For example, in FinTech scenarios, it can recommend relevant financial products based on customer information and product information for financial products like property and casualty insurance and loans. In healthcare scenarios, it can recommend relevant health insurance products based on customer information and product information for these products.

[0003] Currently, when recommending products, multi-category recommendation models are typically used to filter matching products from a product list. However, in real-world scenarios, when customers don't show a clear preference for a particular product, multi-category recommendation models tend to recommend products that are predominant in the product list or have a high historical recommendation success rate. This can result in customers not being interested in the recommended products, and the recommended products may not be comprehensive enough.

[0004] Therefore, how to improve the comprehensiveness of product recommendations has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The main purpose of the embodiments of the present application is to propose a product recommendation method and device, electronic device and storage medium, aiming to improve the comprehensiveness of product recommendations.

[0006] To achieve the above objectives, a first aspect of an embodiment of the present application provides a product recommendation method, the method comprising:

[0007] Obtain product recommendation sample data for candidate products;

[0008] Performing model training on at least two preset original product recommendation binary classification models based on the product recommendation sample data to obtain at least two target product recommendation binary classification models and a model confusion matrix for each of the target product recommendation binary classification models; wherein the model confusion matrix is ​​used to characterize the performance of each of the target product recommendation binary classification models in predicting the recommendation probability of each of the candidate products;

[0009] Performing model weight assignment based on the model confusion matrix to obtain a target model weight matrix;

[0010] Perform product prediction on the target object's user information using multiple target product recommendation binary classification models to obtain the original product recommendation probability;

[0011] Perform product screening based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence;

[0012] Recommend products to the target object based on the target recommended product sequence.

[0013] In some embodiments, the product screening based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence includes:

[0014] For each candidate product, performing weighted summation on the original product recommendation probabilities based on the target model weight matrix to obtain an initial product recommendation probability;

[0015] Normalizing the initial product recommendation probability to obtain a target product recommendation probability;

[0016] Products are screened based on the target product recommendation probability to obtain the target recommended product sequence.

[0017] In some embodiments, for each candidate product, performing weighted summation on the original product recommendation probability based on the target model weight matrix to obtain the initial product recommendation probability includes:

[0018] For each candidate product, weight screening is performed on the target model weight matrix to obtain product weight data of the binary classification model for each target product recommendation;

[0019] Probabilistically screening the original product recommendation probabilities to obtain a product recommendation prediction probability output by each target product recommendation binary classification model;

[0020] The initial product recommendation probability is obtained by performing a weighted summation based on the product weight data and the product recommendation prediction probability.

[0021] In some embodiments, the product screening based on the target product recommendation probability to obtain the target recommended product sequence includes:

[0022] Obtaining a product conversion probability threshold for each candidate product;

[0023] For each candidate product, comparing the target product recommendation probability with the product conversion probability threshold;

[0024] When the target product recommendation probability exceeds the product conversion probability threshold, the candidate product is added to the target recommended product sequence.

[0025] In some embodiments, performing model weight assignment based on the model confusion matrix to obtain a target model weight matrix includes:

[0026] Get the preset benchmark confusion matrix;

[0027] Initialize weights based on the model confusion matrix to obtain an original model weight matrix;

[0028] The original model weight matrix is ​​optimized based on the benchmark confusion matrix and the model confusion matrix to obtain the target model weight matrix.

[0029] In some embodiments, optimizing the original model weight matrix based on the benchmark confusion matrix and the model confusion matrix to obtain the target model weight matrix includes:

[0030] Performing loss calculation on the reference confusion matrix and the model confusion matrix to obtain a confusion matrix loss function;

[0031] Performing gradient calculation on the original model weight matrix based on the confusion matrix loss function to obtain weight gradient data;

[0032] The weight matrix of the original model is updated based on the weight gradient data to obtain the weight matrix of the target model.

[0033] In some embodiments, the performing model training on at least two preset original product recommendation binary classification models based on the product recommendation sample data to obtain at least two target product recommendation binary classification models and a model confusion matrix of each target product recommendation binary classification model includes:

[0034] Dividing the product recommendation sample data into training samples, verification samples, and test samples;

[0035] For each of the original product recommendation binary classification models, model training is performed on the original product recommendation binary classification model based on the training samples to obtain an initial product recommendation binary classification model;

[0036] Performing model validation on the initial product recommendation binary classification model based on the validation sample to obtain model validation data;

[0037] If the model validation data indicates that the initial product recommendation binary classification model has passed the validation, confirming the initial product recommendation binary classification model as the target product recommendation binary classification model;

[0038] A model test is performed on the target product recommendation binary classification model based on the test sample to obtain the model confusion matrix of the target product recommendation binary classification model.

[0039] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a product recommendation device, comprising:

[0040] A sample data acquisition module is used to obtain product recommendation sample data of candidate products;

[0041] A model training module is configured to perform model training on at least two preset original product recommendation binary classification models based on the product recommendation sample data to obtain at least two target product recommendation binary classification models and a model confusion matrix for each of the target product recommendation binary classification models; wherein the model confusion matrix is ​​used to characterize the performance of each of the target product recommendation binary classification models in predicting the recommendation probability of each of the candidate products;

[0042] A model weight allocation module is used to allocate model weights based on the model confusion matrix to obtain a target model weight matrix;

[0043] A recommendation probability prediction module is used to perform product prediction on the user information of the target object using multiple target product recommendation binary classification models to obtain the original product recommendation probability;

[0044] A recommended product screening module is used to screen products based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence;

[0045] The target product recommendation module is used to recommend products to the target object based on the target recommended product sequence.

[0046] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the method described in the first aspect when executing the computer program.

[0047] To achieve the above-mentioned purpose, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect.

[0048] The product recommendation method and device, electronic device, and storage medium proposed in this application obtain product recommendation sample data for candidate products and train multiple original product recommendation binary classification models based on the product recommendation sample data to obtain multiple target product recommendation binary classification models and a model confusion matrix corresponding to each target product recommendation binary classification model. The model confusion matrix is ​​used to characterize the performance of each target product recommendation binary classification model in predicting the recommendation probability of each candidate product, and can intuitively present the prediction effect of each target product recommendation binary classification model on each candidate product, providing a basis for subsequent weight allocation. Furthermore, model weight allocation is performed based on the model confusion matrix to obtain a target model weight matrix. Finally, the trained multiple target product recommendation binary classification models are used to predict products for the user information of the target object, obtaining the original product recommendation probability. Products are then screened based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence. Products are then recommended to the target object based on the target recommended product sequence, rationally utilizing the advantages of each target product recommendation binary classification model and helping to improve the accuracy and comprehensiveness of product recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flowchart of the product recommendation method provided in an embodiment of the present application;

[0050] Figure 2 yes Figure 1 Flowchart of step S102 in FIG.

[0051] Figure 3 yes Figure 1 Flowchart of step S103 in FIG.

[0052] Figure 4 yes Figure 3 Flowchart of step S303 in FIG.

[0053] Figure 5 yes Figure 1 Flowchart of step S105 in FIG.

[0054] Figure 6 yes Figure 5 Flowchart of step S501 in FIG.

[0055] Figure 7 yes Figure 5 Flowchart of step S503 in FIG.

[0056] Figure 8 This is a schematic diagram of the structure of the product recommendation device provided in an embodiment of the present application;

[0057] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0059] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0061] First, let’s analyze some of the terms used in this application:

[0062] Artificial intelligence (AI) is a new technical discipline that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. A branch of computer science, AI seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thinking. It also encompasses the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0063] Natural language processing (NLP): NLP uses computers to process, understand, and apply human languages ​​(such as Chinese and English). A branch of artificial intelligence, NLP is an interdisciplinary field between computer science and linguistics, often referred to as computational linguistics. Natural language processing encompasses grammatical analysis, semantic analysis, and discourse comprehension. Natural language processing is commonly used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis, and opinion mining. It encompasses data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research related to language processing, and linguistics research related to language computing.

[0064] In the field of machine learning, the confusion matrix is ​​an important tool for evaluating the performance of classification models. A confusion matrix is ​​a table that presents the comparison between the classification model's predictions and the true labels in matrix format. The rows of the matrix represent the true categories, and the columns represent the predicted categories. The confusion matrix allows you to visually visualize the number of samples correctly and incorrectly classified by the model, such as true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN). This allows you to calculate metrics such as precision, recall, and F1 score to comprehensively measure the model's classification capabilities.

[0065] Product recommendation is an AI technology that uses specific algorithms and models to filter and recommend products that customers may be interested in based on multiple dimensions of information, including user characteristics, behaviors, preferences, historical data, and product information. The goal of product recommendations is to improve customer discovery efficiency, increase customer purchase intent, and enhance user experience. It also helps companies improve product sales conversion rates and user satisfaction, effectively matching users with products.

[0066] Product recommendations can be applied to multiple scenarios. For example, in FinTech scenarios, relevant financial products can be recommended to customers based on their information and product information of financial products such as property and casualty insurance and loans. In health insurance scenarios, relevant health insurance products can be recommended to customers based on their information and product information of health insurance products.

[0067] Currently, when recommending products, multi-category recommendation models are typically used to filter matching products from a product list. However, in real-world scenarios, when customers don't show a clear preference for a particular product, multi-category recommendation models tend to recommend products that are predominant in the product list or have a high historical recommendation success rate. This can result in customers not being interested in the recommended products, and the recommended products may not be comprehensive enough.

[0068] In addition, for enterprises, recommending more products that have a high proportion in the product list or have a high historical recommendation success rate may lead to an imbalance in product recommendations, resulting in many products failing to have good sales conversions.

[0069] Based on this, the embodiments of the present application provide a product recommendation method and device, an electronic device, and a storage medium, aiming to improve the comprehensiveness of product recommendations.

[0070] The product recommendation method and device, electronic device and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the product recommendation method in the embodiments of the present application is described.

[0071] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0072] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0073] The product recommendation method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The product recommendation method provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, or as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the product recommendation method, etc., but is not limited to the above forms.

[0074] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0075] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of such data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0076] Figure 1 This is an optional flowchart of the product recommendation method provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S106.

[0077] Step S101, obtaining product recommendation sample data of candidate products;

[0078] Step S102: training at least two preset original product recommendation binary classification models based on the product recommendation sample data to obtain at least two target product recommendation binary classification models and a model confusion matrix for each target product recommendation binary classification model; wherein the model confusion matrix is ​​used to represent the performance of each target product recommendation binary classification model in predicting the recommendation probability of each candidate product;

[0079] Step S103, performing model weight assignment based on the model confusion matrix to obtain a target model weight matrix;

[0080] Step S104: using multiple target product recommendation binary classification models to perform product prediction on the target user information to obtain the original product recommendation probability;

[0081] Step S105: Screen products based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence;

[0082] Step S106: recommend products to the target object based on the target recommended product sequence.

[0083] In the steps S101 to S106 shown in the embodiment of the present application, by obtaining product recommendation sample data of candidate products and training multiple original product recommendation binary classification models based on the product recommendation sample data, multiple target product recommendation binary classification models and the model confusion matrix corresponding to each target product recommendation binary classification model are obtained; wherein, the model confusion matrix is ​​used to characterize the performance of each target product recommendation binary classification model in predicting the recommendation probability of each candidate product, and can intuitively present the prediction effect of each target product recommendation binary classification model on each candidate product, providing a basis for subsequent weight allocation. Further, model weight allocation is performed based on the model confusion matrix to obtain a target model weight matrix. Finally, the user information of the target object is predicted by using the trained multiple target product recommendation binary classification models to obtain the original product recommendation probability, and product screening is performed based on the target model weight matrix and the original product recommendation probability to obtain a target recommendation product sequence, and product recommendations are made to the target object based on the target recommendation product sequence, which rationally utilizes the advantages of each target product recommendation binary classification model and helps to improve the accuracy and comprehensiveness of product recommendations.

[0084] In step S101 of some embodiments, the product recommendation sample data is pre-collected product recommendation data in a specific application scenario. Each product recommendation sample is used to recommend a candidate product. The corresponding relationship between the two is: one candidate product corresponds to multiple product recommendation sample data; and one product recommendation sample data corresponds to one candidate product.

[0085] Specifically, the candidate product can be a specific product or a product group including multiple products, which needs to be set according to the actual application scenario, but is not limited thereto, and the number of candidate products is multiple.

[0086] For example, in the financial technology scenario, candidate products may include property insurance products, auto insurance products, loan products, wealth management products, etc.; in the medical insurance scenario, candidate products may include medical insurance products, disease insurance products, life insurance products, etc., but are not limited to these.

[0087] Each product sample recommendation data includes: user information of the sample object, product information of the candidate product, the candidate product and a sample label; wherein the sample label is used to determine whether the candidate product needs to be recommended to the sample object based on the user information and product information.

[0088] In step S102 of some embodiments, the number of preset original product recommendation binary classification models is at least two, and the model structure of each original product recommendation binary classification model is a binary classification model, which is used to output the probability of whether to recommend a certain product based on the input information. It can be assumed that the probability is in the interval [0,0.5], indicating no recommendation, and the probability is in the interval (0.5,1], indicating recommendation.

[0089] For example, a product recommendation sample for a candidate product, "Product A," contains user information for subject 1, product A's information, and a sample label indicating whether Product A is recommended to subject 1. Subject 1's user information and product A's information are fed into the original product recommendation binary classification model for product prediction. The model outputs a recommendation probability for Product A. A probability in the range [0, 0.5] indicates that Product A is not recommended to subject 1, while a probability in the range (0.5, 1] ​​indicates that Product A is recommended to subject 1.

[0090] It should be noted that the model structure of each original product recommendation binary classification model can be the same or different. The original product recommendation binary classification model can adopt a logistic regression model, a decision tree model, a support vector machine model, a neural network model, etc., but is not limited thereto.

[0091] Specifically, each original product recommendation binary classification model needs to be trained using the product recommendation sample data corresponding to each candidate product, so that each target product recommendation binary classification model obtained through training has the ability to predict the recommendation probability of each candidate product (regardless of whether it is accurate or not), and when performing subsequent product predictions, it is necessary to predict the user's recommendation prediction probability for each candidate product.

[0092] See also Figure 2 In some embodiments, step S102 may include but is not limited to steps S201 to S205:

[0093] Step S201: divide the product recommendation sample data into training samples, verification samples, and test samples;

[0094] Step S202: for each original product recommendation binary classification model, the original product recommendation binary classification model is trained based on the training samples to obtain an initial product recommendation binary classification model;

[0095] Step S203: performing model verification on the initial product recommendation binary classification model based on the verification sample to obtain model verification data;

[0096] Step S204: if the model verification data indicates that the initial product recommendation binary classification model has passed the verification, the initial product recommendation binary classification model is confirmed as the target product recommendation binary classification model;

[0097] Step S205 , performing a model test on the target product recommendation binary classification model based on the test samples to obtain a model confusion matrix of the target product recommendation binary classification model.

[0098] In steps S201 to S205 shown in the embodiment of the present application, the product recommendation sample data is divided into training samples, verification samples and test samples, and each original product recommendation binary classification model is trained based on the same training samples to obtain multiple initial product recommendation binary classification models. Then, the initial product recommendation binary classification model is model-verified based on the verification sample to obtain model verification data. When the model verification data indicates that the initial product recommendation binary classification model has passed the verification, the initial product recommendation binary classification model is confirmed as the target product recommendation binary classification model, ensuring the good generalization ability of the model on the verification set. Finally, the target product recommendation binary classification model is model-tested based on the test sample to further verify the accuracy and stability of the model in subsequent applications, and obtain the model confusion matrix of the target product recommendation binary classification model. From data preparation to model training, verification, confirmation and testing, the quality of the model is fully guaranteed, which is conducive to improving the accuracy of product recommendations.

[0099] In step S201 of some embodiments, for each candidate product, the product recommendation sample data corresponding to the candidate product is divided based on a preset division ratio to obtain training samples, verification samples, and test samples. Among them, the training samples are used to train the original product recommendation binary classification model. The verification samples are used to verify the performance of the initial product recommendation binary classification model after preliminary training. The "model training-model verification" method can be used to gradually improve the performance of the model. Alternatively, only one model verification can be used after the model training is completed. The specific needs depend on the specific model structure and are not limited in the embodiments of the present application.

[0100] The test sample is used to test the performance of the target product recommendation binary classification model in predicting the recommendation probability of the candidate product.

[0101] It should be noted that the preset division ratio needs to be determined according to the model structure, and the training samples, validation samples, and test samples are independent of each other to ensure the reliability of subsequent training, validation, and testing steps. For example, the preset division ratio is set to 6:2:2, that is, the number of training samples accounts for 60% of the product recommendation sample data, the number of validation samples accounts for 20% of the product recommendation sample data, and the number of test samples accounts for 20% of the product recommendation sample data.

[0102] In step S202 of some embodiments, for each original product recommendation binary classification model, the original product recommendation binary classification model is trained based on the training samples corresponding to each candidate product to obtain an initial product recommendation binary classification model. During the training process, an optimization algorithm (such as gradient descent) can be used to minimize the prediction loss function, learn the mapping relationship between features and sample labels, and thus learn the underlying patterns in the training samples to obtain the initial product recommendation binary classification model.

[0103] Next, the initial product recommendation binary classification model is validated based on the validation samples corresponding to each candidate product to obtain model validation data. The model validation data is used to characterize the model performance (such as accuracy, AUC, etc.) of the initial product recommendation binary classification model under each candidate product.

[0104] Specifically, if the performance (such as accuracy, AUC, etc.) of the initial product recommendation binary classification model reaches a preset threshold (e.g., 80%) during the prediction process for at least one candidate product, the initial product recommendation binary classification model is considered to have passed verification. For example, if the model's prediction accuracy reaches a preset accuracy threshold (e.g., 90%), or if the model's AUC reaches a preset AUC threshold (e.g., [0.85, 1]), the initial product recommendation binary classification model is considered to have passed verification.

[0105] The verified initial product recommendation binary classification model is confirmed as the target product recommendation binary classification model.

[0106] Finally, the target product recommendation binary classification model is tested based on the test samples corresponding to each candidate product. This yields a model confusion matrix for the target product recommendation binary classification model. This matrix describes the model's prediction results on the test samples and displays the number of true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN) for each candidate product. The confusion matrix can be used to calculate various performance metrics, including accuracy, precision, recall, and F1 score.

[0107] See also Figure 3 In some embodiments, step S103 may include but is not limited to steps S301 to S303:

[0108] Step S301, obtaining a preset reference confusion matrix;

[0109] Step S302, performing weight initialization based on the model confusion matrix to obtain the original model weight matrix;

[0110] Step S303: Optimize the original model weight matrix based on the baseline confusion matrix and the model confusion matrix to obtain the target model weight matrix.

[0111] In steps S301 to S303 shown in the embodiment of the present application, the weights are initialized based on the model confusion matrix to obtain the original model weight matrix, and the original model weight matrix is ​​optimized based on a preset baseline confusion matrix to obtain the target model weight matrix. This can combine the prior knowledge of the baseline confusion matrix to set reasonable weights for multiple target product recommendation binary classification models, which helps to improve the model performance and generalization ability, and thus improve the accuracy of product recommendations.

[0112] In step S301 of some embodiments, the benchmark confusion matrix is ​​an ideal confusion matrix pre-set by business personnel based on actual business needs or ideal model performance, which can reflect the classification performance expected to be achieved by the model in each category.

[0113] It should be noted that because the baseline confusion matrix represents the performance target that the model is expected to achieve, reverse engineering a model based on the baseline confusion matrix is ​​quite difficult. However, adjusting the weights of the existing model based on the baseline confusion matrix and the model confusion matrix is ​​an achievable goal. Therefore, it is necessary to subsequently optimize the original model weight matrix based on the baseline confusion matrix and the model confusion matrix to obtain the target model weight matrix.

[0114] In step S302 of some embodiments, based on the model confusion matrix, the weight parameters of the binary classification model for each target product in predicting the recommendation probability of each candidate product are recommended to obtain the original model weight matrix; wherein the rows of the original model weight matrix represent the binary classification model for each target product recommendation, the columns represent each candidate product, and the original model weight matrix includes multiple original weight elements.

[0115] Specifically: the original weight element can be recorded as W[i][j], where W[i][j] represents the weight of the target product recommendation binary classification model i on the candidate product j.

[0116] In some embodiments, the initial value of each original weight element may be set to 1.

[0117] See also Figure 4 In some embodiments, step S303 may include but is not limited to steps S401 to S403:

[0118] Step S401, performing loss calculation on the reference confusion matrix and the model confusion matrix to obtain a confusion matrix loss function;

[0119] Step S402, performing gradient calculation on the original model weight matrix based on the confusion matrix loss function to obtain weight gradient data;

[0120] Step S403: update the weight matrix of the original model based on the weight gradient data to obtain the target model weight matrix.

[0121] In steps S401 to S403 shown in the embodiment of the present application, by calculating the loss of the baseline confusion matrix and the model confusion matrix, the difference between the two is clarified, and the confusion matrix loss function is obtained to provide direction for subsequent optimization. Then, based on the confusion matrix loss function, the gradient calculation of the original model weight matrix is ​​performed to obtain weight gradient data and determine the weight adjustment amplitude and direction. Finally, based on the weight gradient data, the weight of the original model weight matrix is ​​updated to obtain a better target model weight matrix, which can gradually improve the model performance and prediction accuracy.

[0122] In step S401 of some embodiments, a loss function such as a mean square error loss function and a cross entropy loss function can be used to calculate the loss of the baseline confusion matrix and the model confusion matrix, thereby measuring the difference between the baseline confusion matrix and the model confusion matrix and obtaining a confusion matrix loss function.

[0123] Furthermore, by calculating the gradient of the confusion matrix loss function for each original weight element W[i][j] in the original model weight matrix, the weight adjustment direction is guided to obtain the weight gradient data.

[0124] Therefore, the weight matrix of the original model is updated according to the weight gradient data to obtain the weight matrix of the target model.

[0125] The target model weight matrix includes multiple target weight elements.

[0126] Specifically, the original weight element is updated based on the weight gradient data to obtain the target weight element, see formula (1);

[0127]

[0128] Among them, W ′ [i,j] is the target weight element, W[i,j] is the original weight element, η is the learning rate of the target product recommendation binary classification model, is the weight gradient data.

[0129] For example, suppose there are two candidate products, namely Product 1 and Product 2; there are three target product recommendation binary classification models, namely Model 1, Model 2 and Model 3;

[0130] The model confusion matrix of each model is shown in Table 1;

[0131] Model 1 Model 2 Model 3 Product 1 prediction is accurate 90％ 80％ 85％ Product 1 Forecast Error 10％ 20％ 15％ Product 2 prediction is accurate 20％ 10％ 15％ Product 2 prediction error 80％ 90％ 85％

[0132] Table 1

[0133] As can be seen from Table 1, Model 1 performs better on Product 1, Model 2 performs better on Product 2, and Model 3 has a relatively balanced overall performance;

[0134] The preset baseline model confusion matrix is ​​shown in Table 2:

[0135] Model Accurate product forecasts 95％ Product forecast errors 5％

[0136] Table 2

[0137] Based on the model confusion matrix, an original model weight matrix is ​​established, and the initial value of each original weight element is 1.

[0138] It can be seen that the original weight element W 原始 [1,1]=1, that is, the weight of the output result of model 1 when predicting product 1 is 1, and the rest are similar.

[0139] Next, the loss is calculated based on the baseline model confusion matrix and the model confusion matrix to obtain the confusion matrix loss function. The gradient of the original model weight matrix is ​​calculated based on the confusion matrix loss function to obtain the weight gradient data. Then, the weight of the original model weight matrix is ​​updated based on the weight gradient data to obtain the target model weight matrix. The target model weight matrix is

[0140] It should be noted that the above examples are for illustration only and are not intended to limit the embodiments of the present application.

[0141] In some embodiments, when the product recommendation method provided in the embodiments of the present application is subsequently put into actual business scenarios for use, the target model weight matrix can also be adjusted according to specific business feedback to meet the needs of the business scenarios.

[0142] In step S104 of some embodiments, for each target product recommendation binary classification model, it is necessary to predict the target user's recommendation prediction probability under each candidate product, that is: for each candidate product, the product information corresponding to the candidate product and the user information of the target object are input into the target product recommendation binary classification model for prediction, and the recommendation prediction probabilities corresponding to multiple candidate products are obtained, thereby forming an original product recommendation probability.

[0143] Specifically, the original product recommendation probability is a matrix, where the rows represent each target product recommendation binary classification model, the columns represent each candidate product, and the corresponding values ​​are the recommendation prediction probabilities corresponding to a candidate product predicted by a target product recommendation binary classification model.

[0144] See also Figure 5 In some embodiments, step S105 may also include but is not limited to steps S501 to S503:

[0145] Step S501: For each candidate product, perform weighted summation on the original product recommendation probability based on the target model weight matrix to obtain the initial product recommendation probability;

[0146] Step S502: normalize the initial product recommendation probability to obtain the target product recommendation probability;

[0147] Step S503: Screen products based on the target product recommendation probability to obtain a target recommended product sequence.

[0148] In steps S501 to S503, as shown in the embodiment of this application, the initial product recommendation probability for each candidate product is obtained by weighted summation of the original product recommendation probabilities based on the target model weight matrix, thereby strengthening the influence of each model on the recommendation. Next, the initial product recommendation probabilities are normalized to obtain the target product recommendation probability, and products are screened based on the normalized target product recommendation probabilities to obtain the target recommended product sequence. This improves the scientific nature and effectiveness of product recommendations, avoids the tendency to recommend products that have a high proportion in the product list or have a high historical recommendation success rate, and can improve the comprehensiveness of product recommendations.

[0149] In step S501 of some embodiments, for each candidate product, the recommendation probability output by each target product recommendation binary classification model is determined from the original product recommendation probability, and the recommendation probabilities output by each target product recommendation binary classification model are weighted and summed based on the target model weight matrix. This can integrate the prediction performance of each target product recommendation binary classification model, helping to improve the accuracy of product recommendation prediction. Thus, an initial product recommendation probability for a candidate product is obtained for the target object.

[0150] In step S502 of some embodiments, for the target object, the initial product recommendation probability includes the product recommendation probability corresponding to each candidate product.

[0151] See also Figure 6 In some embodiments, step S501 includes but is not limited to steps S601 to S603:

[0152] Step S601: For each candidate product, weight screening is performed on the target model weight matrix to obtain product weight data of the binary classification model for each target product recommendation;

[0153] Step S602: Probabilistically screen the original product recommendation probabilities to obtain the product recommendation prediction probability output by the binary classification model for each target product recommendation;

[0154] Step S603: Perform weighted summation based on the product weight data and the product recommendation prediction probability to obtain an initial product recommendation probability.

[0155] In steps S601 to S603, as shown in the embodiment of the present application, for each candidate product, the product weight data of each target product recommendation binary classification model is determined based on the target model weight matrix, and the product recommendation prediction probability output by each target product recommendation binary classification model is determined from the original product recommendation probability. Finally, a weighted summation is performed based on the product weight data and recommendation prediction probability of each target product recommendation binary classification model, which can integrate the prediction performance of each target product recommendation binary classification model and help improve the accuracy of product recommendation predictions.

[0156] For example, there is a user 1, 5 target product recommendation binary classification models, namely Model A, Model B, Model C, Model D, Model E, and 4 candidate products, namely Candidate Product A, Candidate Product B, Candidate Product C, Candidate Product D;

[0157] The original product recommendation probability obtained is shown in Table 3:

[0158] Model A Model B Model C Model D Model E Candidate Product A 0.7 0.85 0.9 0.4 0.2 Candidate Product B 0.5 0.2 0.5 0.9 0.6 Candidate Product C 0.9 0.8 0.2 0.6 0.5 Candidate Product D 0.3 0.6 0.7 0.5 0.9

[0159] Table 3

[0160] The target model weight matrix is:

[0161] Among them, the target weight element W[i][j] in the target model weight matrix represents the weight of the target product recommendation binary classification model i on the candidate product j.

[0162] Therefore, the initial product recommendation probability corresponding to each candidate product is obtained:

[0163] The initial product recommendation probability corresponding to candidate product A is:

[0164] PA=0.6*0.7+0.9*0.85+1*0.9+0.4*0.4+0.3*0.2=2.305;

[0165] The initial product recommendation probability corresponding to candidate product B is:

[0166] PB=0.3*0.5+0.1*0.2+0.3*0.5+1*0.9+0.5*0.6=1.52;

[0167] The initial product recommendation probability corresponding to candidate product C is:

[0168] PC=1*0.9+0.8*0.8+0.1*0.2+0.5*0.6+0.6*0.5=2.16;

[0169] The initial product recommendation probability corresponding to candidate product D is:

[0170] PD=0.1*0.3+0.3*0.6+0.3*0.7+0.2*0.5+1*0.9=1.42.

[0171] Finally, the initial product recommendation probability is summarized as [2.305, 1.52, 2.16, 1.42].

[0172] In step S502 of some embodiments, the sigmoid function is used to normalize the initial product recommendation probability, thereby mapping the initial product recommendation probability to the (0,1) interval, and the target product recommendation probability is output for binary classification. If the product conversion probability threshold corresponding to the candidate product is exceeded, the candidate product is recommended.

[0173] See also Figure 7 In some embodiments, step S503 may include but is not limited to steps S701 to S703:

[0174] Step S701, obtaining a product conversion probability threshold for each candidate product;

[0175] Step S702: for each candidate product, compare the target product recommendation probability with the product conversion probability threshold;

[0176] Step S703: When the target product recommendation probability exceeds the product conversion probability threshold, the candidate product is added to the target recommended product sequence.

[0177] In steps S701 to S703 shown in the embodiment of the present application, the product conversion probability threshold of each candidate product is obtained; for each candidate product, the target product recommendation probability is compared with the product conversion probability threshold. When the target product recommendation probability exceeds the product conversion probability threshold, the candidate product is added to the target recommended product sequence, which can effectively filter out products with low conversion probability and improve the quality of product recommendations.

[0178] It should be noted that the product conversion probability threshold of each candidate product is set according to actual business needs. For example, for products with a higher proportion in the product list or a higher historical recommendation success rate, the corresponding product conversion probability threshold needs to be set higher, such as 0.6, 0.7, etc., while for products with a lower proportion in the product list or a lower historical recommendation success rate, the corresponding product conversion probability threshold needs to be set lower, such as 0.3, 0.25, etc., or even 0.05.

[0179] Furthermore, when the product recommendation method provided in the embodiment of the present application is subsequently put into use in actual business scenarios, the product conversion probability threshold may be adjusted based on specific business feedback.

[0180] It is understandable that the method of setting the product conversion probability threshold can effectively balance the accuracy and comprehensiveness of product recommendations. For products with high proportions or high success rates, a higher threshold (such as 0.6) can ensure the quality of recommendations and avoid invalid exposure; while for products with low proportions or low success rates, a lower threshold (such as 0.05) can increase the display opportunities of long-tail products, promote new product incubation and market diversity. By dynamically adapting the product conversion probability threshold of different products according to business scenarios, the overall recommendation efficiency is improved and the comprehensiveness of product recommendations is optimized.

[0181] In some embodiments, for each candidate product, the target product recommendation probability is compared with the product conversion probability threshold. If the target product recommendation probability exceeds the product conversion probability threshold, the product can be recommended to the target object, and thus the candidate product is added to the target recommended product sequence.

[0182] If the target product recommendation probability does not exceed the product conversion probability threshold, the product cannot be recommended to the target object.

[0183] In step S106 of some embodiments, after the final target recommended product sequence is generated, the target recommended product sequence is recommended to the target object through a preset channel, so that the target object will have the idea of ​​purchasing / subscribing when learning about the products in the target recommended product sequence.

[0184] The preset channels may be recommended on a recommendation page of an application, or may be recommended to a target object by making an outbound call, or may be recommended to a target object by using text messages, without limitation thereto.

[0185] See also Figure 8 The present application also provides a product recommendation device that can implement the above-mentioned product recommendation method. The device includes:

[0186] The sample data acquisition module 801 is used to acquire product recommendation sample data of candidate products;

[0187] A model training module 802 is configured to perform model training on at least two preset original product recommendation binary classification models based on the product recommendation sample data, thereby obtaining at least two target product recommendation binary classification models and a model confusion matrix for each target product recommendation binary classification model; wherein the model confusion matrix is ​​used to represent the prediction performance of each target product recommendation binary classification model in predicting the recommendation probability of each candidate product;

[0188] A model weight allocation module 803 is used to allocate model weights based on the model confusion matrix to obtain a target model weight matrix;

[0189] The recommendation probability prediction module 804 is used to perform product prediction on the user information of the target object using multiple target product recommendation binary classification models to obtain the original product recommendation probability;

[0190] The recommended product screening module 805 is used to screen products based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence;

[0191] The target product recommendation module 806 is used to recommend products to the target object based on the target recommended product sequence.

[0192] The specific implementation of the product recommendation device is basically the same as the specific embodiment of the above-mentioned product recommendation method, and will not be repeated here.

[0193] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the product recommendation method when executing the computer program. The electronic device can be any smart terminal, such as a tablet computer or an in-vehicle computer.

[0194] See also Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0195] The processor 901 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0196] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the product recommendation method of the embodiments of this application;

[0197] Input / output interface 903, used to implement information input and output;

[0198] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0199] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0200] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0201] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned product recommendation method is implemented.

[0202] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0203] The product recommendation method and device, electronic device and storage medium provided in the embodiments of the present application obtain product recommendation sample data of candidate products, and train multiple original product recommendation binary classification models based on the product recommendation sample data to obtain multiple target product recommendation binary classification models and the model confusion matrix corresponding to each target product recommendation binary classification model; wherein, the model confusion matrix is ​​used to characterize the performance of each target product recommendation binary classification model in predicting the recommendation probability of each candidate product, and can intuitively present the prediction effect of each target product recommendation binary classification model on each candidate product, providing a basis for subsequent weight allocation. Further, model weight allocation is performed based on the model confusion matrix to obtain a target model weight matrix. Finally, the user information of the target object is predicted using the trained multiple target product recommendation binary classification models to obtain the original product recommendation probability, and product screening is performed based on the target model weight matrix and the original product recommendation probability to obtain a target recommendation product sequence, and product recommendations are made to the target object based on the target recommendation product sequence, which rationally utilizes the advantages of each target product recommendation binary classification model and helps to improve the accuracy and comprehensiveness of product recommendations.

[0204] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0205] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0206] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0207] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0208] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0209] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0210] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0211] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0212] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0213] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0214] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A product recommendation method, characterized in that: The method comprises: Obtain product recommendation sample data for candidate products; Performing model training on at least two preset original product recommendation binary classification models based on the product recommendation sample data to obtain at least two target product recommendation binary classification models and a model confusion matrix for each of the target product recommendation binary classification models; wherein the model confusion matrix is ​​used to characterize the performance of each of the target product recommendation binary classification models in predicting the recommendation probability of each of the candidate products; Performing model weight assignment based on the model confusion matrix to obtain a target model weight matrix; Perform product prediction on the target object's user information using multiple target product recommendation binary classification models to obtain the original product recommendation probability; Perform product screening based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence; Recommend products to the target object based on the target recommended product sequence.

2. The method according to claim 1, characterized in that The product screening based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence includes: For each candidate product, performing weighted summation on the original product recommendation probabilities based on the target model weight matrix to obtain an initial product recommendation probability; Normalizing the initial product recommendation probability to obtain a target product recommendation probability; Products are screened based on the target product recommendation probability to obtain the target recommended product sequence.

3. The method according to claim 2, characterized in that For each candidate product, performing weighted summation on the original product recommendation probability based on the target model weight matrix to obtain an initial product recommendation probability includes: For each candidate product, weight screening is performed on the target model weight matrix to obtain product weight data of the binary classification model for each target product recommendation; Probabilistically screening the original product recommendation probabilities to obtain a product recommendation prediction probability output by each target product recommendation binary classification model; The initial product recommendation probability is obtained by performing a weighted summation based on the product weight data and the product recommendation prediction probability.

4. The method according to claim 2, characterized in that The product screening based on the target product recommendation probability to obtain the target recommended product sequence includes: Obtaining a product conversion probability threshold for each candidate product; For each candidate product, comparing the target product recommendation probability with the product conversion probability threshold; When the target product recommendation probability exceeds the product conversion probability threshold, the candidate product is added to the target recommended product sequence.

5. The method according to claim 1, wherein The model weight distribution is performed based on the model confusion matrix to obtain a target model weight matrix, including: Get the preset benchmark confusion matrix; Initialize weights based on the model confusion matrix to obtain an original model weight matrix; The original model weight matrix is ​​optimized based on the benchmark confusion matrix and the model confusion matrix to obtain the target model weight matrix.

6. The method according to claim 5, characterized in that The optimizing the original model weight matrix based on the benchmark confusion matrix and the model confusion matrix to obtain the target model weight matrix includes: Performing loss calculation on the reference confusion matrix and the model confusion matrix to obtain a confusion matrix loss function; Performing gradient calculation on the original model weight matrix based on the confusion matrix loss function to obtain weight gradient data; The weight matrix of the original model is updated based on the weight gradient data to obtain the weight matrix of the target model.

7. The method according to any one of claims 1 to 6, characterized in that The step of training at least two preset original product recommendation binary classification models based on the product recommendation sample data to obtain at least two target product recommendation binary classification models and a model confusion matrix of each target product recommendation binary classification model includes: Dividing the product recommendation sample data into training samples, verification samples, and test samples; For each of the original product recommendation binary classification models, model training is performed on the original product recommendation binary classification model based on the training samples to obtain an initial product recommendation binary classification model; Performing model validation on the initial product recommendation binary classification model based on the validation sample to obtain model validation data; If the model validation data indicates that the initial product recommendation binary classification model has passed the validation, confirming the initial product recommendation binary classification model as the target product recommendation binary classification model; A model test is performed on the target product recommendation binary classification model based on the test sample to obtain the model confusion matrix of the target product recommendation binary classification model.

8. A product recommendation device, characterized in that: The device comprises: A sample data acquisition module is used to obtain product recommendation sample data marked with candidate products; A model training module is configured to perform model training on at least two preset original product recommendation binary classification models based on the product recommendation sample data to obtain at least two target product recommendation binary classification models and a model confusion matrix for each of the target product recommendation binary classification models; wherein the model confusion matrix is ​​used to characterize the performance of each of the target product recommendation binary classification models in predicting the recommendation probability of each of the candidate products; A model weight allocation module is used to allocate model weights based on the model confusion matrix to obtain a target model weight matrix; A recommendation probability prediction module is used to perform product prediction on the user information of the target object using multiple target product recommendation binary classification models to obtain the original product recommendation probability; A recommended product screening module is used to screen products based on the target model weight matrix and the original product recommendation probability to obtain a target recommended product sequence; The target product recommendation module is used to recommend products to the target object based on the target recommended product sequence.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

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

Images