Artificial intelligence-based intelligent recommendation method, device, electronic device, and medium

Abstract

The present invention relates to the field of artificial intelligence technology, and provides an artificial intelligence-based intelligent recommendation method, device, electronic device and medium, the method comprising: collecting a first data set corresponding to the previous cycle of the current cycle from multiple data sources at preset intervals and preprocessing the data to obtain a second data set; cutting the second data set to obtain a target data set; inputting each sub-target data set in the target data set into a plurality of pre-trained indicator calculation models to obtain multiple indicator values ​​for each customer; and determining a target recommendation scheme based on the multiple indicator values ​​of each customer and the product portraits of all types of recommended products. The present invention calculates multiple indicator values ​​for each customer through multiple indicator calculation models, and when determining the target recommendation scheme, takes into account the multiple indicators of each customer, thereby improving the accuracy of the target recommendation scheme.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based intelligent recommendation method, device, electronic device, and medium. Background Art

[0002] Traditional intelligent recommendation is based on a large-scale data model for intelligent recommendation services. It requires collecting data sets from data sources, using a single model to calculate and verify the collected data sets, and finally drawing model conclusions to make some intelligent or non-intelligent recommendations on data and guide customer behavior.

[0003] However, existing technologies use a single model to calculate indicators and make product recommendations based on the calculated single indicator value. This does not take into account the multiple dimensions of customers and the variability of financial data. At the same time, the data results need to be verified, resulting in low efficiency and accuracy in product recommendations.

[0004] Therefore, it is necessary to propose a method that can recommend products quickly and accurately. Summary of the Invention

[0005] In view of the above content, it is necessary to propose an intelligent recommendation method, device, electronic device and medium based on artificial intelligence, which calculates multiple indicator values ​​for each customer through multiple indicator calculation models. When determining the target recommendation plan, multiple indicators of each customer are taken into consideration, thereby improving the accuracy of the target recommendation plan.

[0006] A first aspect of the present invention provides an artificial intelligence-based intelligent recommendation method, the method comprising:

[0007] Parse the received intelligent recommendation request and obtain multiple data sources;

[0008] Collecting a first data set corresponding to a previous period of a current period from the multiple data sources at every preset period, and preprocessing each data in the first data set to obtain a second data set;

[0009] Cutting the second data set to obtain a target data set;

[0010] Inputting each target sub-dataset in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer;

[0011] Based on the multiple indicator values ​​of each customer and the product profiles of all types of recommended products, a target recommendation plan for the corresponding customer is determined.

[0012] Optionally, the segmenting the second data set to obtain a target data set includes:

[0013] The second data set is classified according to the intelligent recommendation task in the intelligent recommendation request, and the classified second data set is cut according to a preset first cutting rule to obtain a second sub-data set; the customer numbers of all customers in each of the second sub-data sets are identified, and each of the second sub-data sets is cut in the order of the last digits of the customer numbers of all customers according to the preset second cutting rule to obtain a third sub-data set corresponding to each of the second sub-data sets; the third sub-data sets are merged, and the merged third sub-data set is determined as the target data set.

[0014] Optionally, inputting each target sub-dataset in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer includes:

[0015] Inputting each target sub-dataset into a pre-trained response rate calculation model to obtain the response rate of each customer;

[0016] Input each target sub-dataset into a pre-trained willingness calculation model to obtain the willingness of each customer;

[0017] Input each target sub-dataset into a pre-trained behavior score calculation model to obtain a behavior score for each customer;

[0018] Input each target sub-dataset into a pre-trained credit score calculation model to obtain the credit score of each customer;

[0019] Input each target sub-dataset into a pre-trained customer profile calculation model to obtain a customer profile for each customer;

[0020] The response rate, willingness, behavior score, credit value and customer portrait of each customer are determined as multiple indicator values ​​for each customer.

[0021] Optionally, determining a target recommendation plan for each customer based on the multiple indicator values ​​of each customer and the product profiles of all recommended products includes:

[0022] Determine each customer's willingness to close a deal based on each customer's response rate and corresponding willingness;

[0023] Determine the target recommended products for each customer based on their behavior score, credit score, customer profile, and product profiles of all recommended products;

[0024] Based on each customer's willingness to make a deal and the target recommended products for the corresponding customer, determine the target recommendation plan for the corresponding customer.

[0025] Optionally, determining each customer's transaction willingness based on each customer's response rate and corresponding willingness includes:

[0026] Calculate the product of each customer's response rate and a preset first weight value to obtain a first score;

[0027] Calculate the product of each customer's willingness and a preset second weight value to obtain a second score;

[0028] The sum of the first score and the second score is calculated to obtain each customer's transaction willingness.

[0029] Optionally, determining the target recommended product for the corresponding customer based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products includes:

[0030] Construct a decision tree based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products. Each layer of the decision tree includes attribute nodes and rule nodes corresponding to the attribute nodes. The attribute nodes refer to attribute nodes corresponding to multiple attributes obtained by combining each customer's behavior score, credit score, and customer profile.

[0031] Locating a target attribute node corresponding to the attribute of each customer in the decision tree to obtain a first target rule node corresponding to the target attribute node in a target layer;

[0032] If the first target rule node of the target layer is a subtree, continue locating according to the attribute node of the next layer of the subtree to obtain a second target rule node corresponding to the attribute node of the next layer, until the target rule node is determined to be a leaf node, and then determine the product of the target rule node as the target recommended product for the corresponding customer; or

[0033] If the first target rule node of the target layer is a leaf node, the product of the first target rule node is determined to be the target recommended product for the corresponding customer.

[0034] Optionally, constructing a decision tree based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products includes:

[0035] Determine attribute nodes based on each customer's behavior score, credit score, and customer profile;

[0036] Traverse all attribute nodes and product portraits of all recommended products of the aforementioned categories to obtain all recommended products of the categories corresponding to each attribute node as rule nodes;

[0037] The attribute with the largest weight is used as the root node in the attribute nodes;

[0038] If the rule node includes a category of recommended products, determining that the rule node is a leaf node;

[0039] If the rule node includes at least two categories of recommended products, determining that the rule node is a subtree;

[0040] The attribute with the largest weight except the upper layer is selected as the attribute node of the next layer of the subtree.

[0041] A second aspect of the present invention provides an artificial intelligence-based intelligent recommendation device, comprising:

[0042] The parsing module is used to parse the received intelligent recommendation request and obtain multiple data sources;

[0043] an acquisition module, configured to acquire, from the plurality of data sources at predetermined intervals, a first data set corresponding to a period preceding a current period, and preprocess each data in the first data set to obtain a second data set;

[0044] a cutting module, configured to cut the second data set to obtain a target data set;

[0045] An input module, configured to input each target sub-dataset in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer;

[0046] The determination module is used to determine the target recommendation plan for the corresponding customer based on the multiple indicator values ​​of each customer and the product portraits of all types of recommended products.

[0047] A third aspect of the present invention provides an electronic device, comprising a processor and a memory, wherein the processor is configured to implement the artificial intelligence-based intelligent recommendation method when executing a computer program stored in the memory.

[0048] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the artificial intelligence-based intelligent recommendation method when executed by a processor.

[0049] In summary, the artificial intelligence-based intelligent recommendation method, device, electronic device and medium described in the present invention collects a first data set from multiple data sources according to a preset period, limits the data range of the first data set, and ensures the validity of the data collected in the first data set. The collected first data is pre-processed and cut to avoid the interference of errors and duplicate data, solves the problem of low efficiency and accuracy of subsequent data processing caused by inconsistent formats, improves the efficiency of product recommendations, and at the same time, the obtained target data set is distributed and operated, which improves the operating efficiency of the server and thus improves the efficiency of obtaining the target data set. Each of the target sub-datasets is input into a plurality of pre-trained indicator calculation models, and the multiple indicator values ​​of each customer are calculated. Based on the multiple indicator values ​​of each customer and the product portraits of all types of recommended products, the target recommendation plan for each customer is determined, and when determining the target recommendation plan for each customer, the multiple indicators of each customer are taken into consideration, thereby improving the accuracy of the target recommendation plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flowchart of an artificial intelligence-based intelligent recommendation method provided in Example 1 of the present invention.

[0051] Figure 2 This is a structural diagram of an artificial intelligence-based intelligent recommendation device provided in Example 2 of the present invention.

[0052] Figure 3 It is a structural diagram of an electronic device provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0053] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features therein can be combined with each other without conflict.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0055] Example 1

[0056] Figure 1 This is a flowchart of an artificial intelligence-based intelligent recommendation method provided in Example 1 of the present invention.

[0057] In this embodiment, the artificial intelligence-based intelligent recommendation method can be applied to electronic devices. For electronic devices that need to perform artificial intelligence-based intelligent recommendations, the artificial intelligence-based intelligent recommendation function provided by the method of the present invention can be directly integrated into the electronic device, or run in the electronic device in the form of a software development kit (SDK).

[0058] Embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0059] 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, machine learning, and deep learning.

[0060] like Figure 1 As shown, the artificial intelligence-based intelligent recommendation method specifically includes the following steps. According to different needs, the order of the steps in the flowchart can be changed, and some can be omitted.

[0061] S11, parse the received intelligent recommendation request and obtain multiple data sources.

[0062] In this embodiment, in the financial field, due to the wide variety of products, customers cannot quickly determine the products they want to buy and need the server to recommend products. Therefore, the client initiates an intelligent recommendation request to the server. Specifically, the client can be a smart phone, IPAD or other existing smart devices, and the server can be an intelligent recommendation subsystem. During the product recommendation process, the client can send an intelligent recommendation request to the intelligent recommendation subsystem, and the intelligent recommendation subsystem is used to receive the intelligent recommendation request sent by the client.

[0063] In this embodiment, when the intelligent recommendation subsystem receives an intelligent recommendation request, it parses the message information in the intelligent recommendation request, and obtains multiple data sources through the calling interface according to the calling interface in the message information, wherein different data sources correspond to different calling interfaces. Specifically, the data sources may include: financial institution platforms, third-party platforms that have an association with the recommended products of the financial institution platforms, wherein the third-party platforms may include third-party payment platforms, bank transaction platforms, shopping platforms, third-party consumption platforms, etc.

[0064] Specifically, the intelligent recommendation request includes an intelligent recommendation task and a customer group corresponding to the intelligent recommendation task.

[0065] S12: collecting a first data set corresponding to a previous period of a current period from the multiple data sources at every preset period, and preprocessing each data in the first data set to obtain a second data set.

[0066] In this embodiment, considering the variability of data in the financial field, a data collection period can be preset. For example, the preset period can be set to 24 hours. If the current collection time is 9 o'clock, the data set within 24 hours before 9 o'clock is collected.

[0067] Specifically, the variability of data in the financial field refers to the rapid changes in customer data. For example, on day N, the customer's credit score is 9.5 points. On day N+1, the customer's credit score may become 2 points because the customer did not pay off the credit card in time or owed debts.

[0068] In an optional embodiment, preprocessing each data in the first data set to obtain the second data set includes:

[0069] Performing data cleaning and data deduplication on each data in the first data set to obtain a preprocessed first data set;

[0070] The format of each data in the preprocessed first data set is converted according to a preset format conversion rule, and each data after the format conversion is merged to obtain a second data set.

[0071] Specifically, the data cleaning includes one or more of the following methods: cleaning of missing values, format content, logical errors, and non-required data; and data deduplication refers to removing identical data.

[0072] Specifically, data conversion rules may be preset, for example, converting the data format of all date data into XXXX / XX / XX.

[0073] In this embodiment, the first data set is collected from multiple data sources according to a preset period, and the data range of the first data set is restricted, thereby ensuring the validity of the collected data of the first data set.

[0074] In this embodiment, by performing data cleaning, data deduplication and format conversion on each data in the first data set, the interference of erroneous and duplicate data is avoided, and at the same time, the problem of low efficiency and accuracy of subsequent data processing caused by inconsistent formats is solved, thereby improving the product recommendation efficiency.

[0075] S13: Cut the second data set to obtain a target data set.

[0076] In this embodiment, there is a large amount of financial data. In order to ensure the efficiency of subsequent data recommendation and improve the operating efficiency of the server, the second data set is segmented and the obtained target data set is distributed.

[0077] In an optional embodiment, segmenting the second data set to obtain a target data set includes:

[0078] The second data set is segmented according to a preset segmentation rule to obtain a target data set.

[0079] In other optional embodiments, the step of segmenting the second data set according to a preset segmentation rule to obtain a target data set includes:

[0080] Classifying the second data set according to the intelligent recommendation task in the intelligent recommendation request, and segmenting the classified second data set according to a preset first segmentation rule to obtain a second sub-data set;

[0081] Identifying the customer numbers of all customers in each of the second sub-datasets, and segmenting each of the second sub-datasets in the order of the last digits of the customer numbers of all customers according to a preset second segmentation rule to obtain a third sub-dataset corresponding to each of the second sub-datasets;

[0082] The third sub-dataset is merged, and the merged third sub-dataset is determined as the target dataset.

[0083] In this embodiment, the preset first cutting rule is set based on the task volume of the intelligent recommendation task in the intelligent recommendation request, and the preset second cutting rule is set based on the customer volume of all customers in the intelligent recommendation task.

[0084] For example, if the smart recommendation request contains 800,000 smart recommendation tasks, the 800,000 smart recommendation tasks are divided into 8 second sub-data sets according to the preset first cutting rule, and each second sub-data set is cut according to the preset second cutting rule in the order of the last digits of the customer numbers of all customers to obtain the target data set corresponding to each second sub-data set, wherein the last digit refers to a serial number that is different from other customers.

[0085] In this embodiment, by segmenting the second data set and performing distributed operation on the obtained target data set, the operation efficiency of the server can be improved, thereby improving the efficiency of obtaining the target data set.

[0086] S14, inputting each of the target sub-datasets in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer.

[0087] In this embodiment, the target data set includes multiple target sub-data sets. In order to better perform product recommendations, multiple indicator calculation models can be pre-trained, and multiple indicators of each customer in each of the target sub-data sets can be calculated separately through the preset multiple indicator calculation models.

[0088] In an optional embodiment, inputting each target sub-dataset in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer includes:

[0089] Inputting each target sub-dataset into a pre-trained response rate calculation model to obtain the response rate of each customer;

[0090] Input each target sub-dataset into a pre-trained willingness calculation model to obtain the willingness of each customer;

[0091] Input each target sub-dataset into a pre-trained behavior score calculation model to obtain a behavior score for each customer;

[0092] Input each target sub-dataset into a pre-trained credit score calculation model to obtain the credit score of each customer;

[0093] Input each target sub-dataset into a pre-trained customer profile calculation model to obtain a customer profile for each customer;

[0094] The response rate, willingness, behavior score, credit value and customer portrait of each customer are determined as multiple indicator values ​​for each customer.

[0095] In this embodiment, the indicator calculation model may include other indicator calculation models such as a response rate calculation model, a willingness calculation model, a behavior scoring calculation model, a credit value calculation model and a customer portrait calculation model. The indicator calculation model is pre-trained according to the recommendation needs, wherein the training process of the multiple indicator calculation models is the same, the difference lies in the different input sample data sets. Specifically, for the training of the response rate calculation model, the input sample data set is: feature values ​​such as customer visits to products and purchase records; for the training of the willingness calculation model, the input sample data set is: feature values ​​such as customer scores and comments on each product; for the training of the behavior scoring calculation model, the input sample data set is: behavioral features such as average balance, number of overdue payments, and credit limit; for the training of the credit value calculation model, the input sample data set is: feature values ​​such as customer loan repayment records, credit information, and credit card usage records; for the training of the customer portrait calculation model, the input sample data set is: feature values ​​such as customer age, gender, and product preferences.

[0096] Specifically, the training process of the indicator calculation model includes: dividing a training set and a test set from the sample data set; inputting the training set into a preset neural network for training to obtain an indicator calculation model; inputting the test set into the indicator calculation model for testing, and calculating the test pass rate; if the test pass rate is greater than or equal to a preset pass rate threshold, determining that the training of the indicator calculation model is completed; if the test pass rate is less than the preset pass rate threshold, increasing the number of training sets and re-training the indicator calculation model.

[0097] S15, determining a target recommendation plan for the corresponding customer based on the multiple indicator values ​​of each customer and the product portraits of all types of recommended products.

[0098] In this embodiment, due to the wide variety of product types in the financial industry, when determining the target recommendation plan, it is necessary to obtain the product information of all recommended products, and input the product information of all products into a pre-trained classification model to obtain the product labels of each type of recommended products, and construct a product portrait of each type of product based on the product labels of each type of recommended products, thereby improving the recommendation efficiency of subsequent target recommendation plans.

[0099] In an optional embodiment, determining a target recommendation plan for each customer based on the multiple indicator values ​​of each customer and the product profiles of all recommended products includes:

[0100] Determine each customer's willingness to close a deal based on each customer's response rate and corresponding willingness;

[0101] Determine the target recommended products for each customer based on their behavior score, credit score, customer profile, and product profiles of all recommended products;

[0102] Based on each customer's willingness to make a deal and the target recommended products for the corresponding customer, determine the target recommendation plan for the corresponding customer.

[0103] In this embodiment, the response rate refers to the customer's response to the product, that is, whether the customer opens the product link, etc.; the willingness refers to the customer's willingness to the product; and the transaction willingness refers to the customer's willingness to complete the transaction.

[0104] Furthermore, determining each customer's transaction willingness based on each customer's response rate and corresponding willingness includes:

[0105] Calculate the product of each customer's response rate and a preset first weight value to obtain a first score;

[0106] Calculate the product of each customer's willingness and a preset second weight value to obtain a second score;

[0107] The sum of the first score and the second score is calculated to obtain each customer's transaction willingness.

[0108] In this embodiment, when calculating each customer's willingness to complete a transaction, taking into account the differences in the impact of response rate and willingness on the willingness to complete a transaction, a first weight value is set in advance for the response rate, and a second weight value is set in advance for the willingness. The first weight value and the second weight value can be set equal or unequal according to the recommended task situation, and this embodiment does not impose any restrictions on this.

[0109] Furthermore, the target recommended products for each customer are determined based on the behavior score, credit score, customer profile, and product profiles of all recommended products of each customer, including:

[0110] Construct a decision tree based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products. Each layer of the decision tree includes attribute nodes and rule nodes corresponding to the attribute nodes. The attribute nodes refer to attribute nodes corresponding to multiple attributes obtained by combining each customer's behavior score, credit score, and customer profile.

[0111] Locating a target attribute node corresponding to the attribute of each customer in the decision tree to obtain a first target rule node corresponding to the target attribute node in a target layer;

[0112] If the first target rule node of the target layer is a subtree, continue locating according to the attribute node of the next layer of the subtree to obtain a second target rule node corresponding to the attribute node of the next layer, until the target rule node is determined to be a leaf node, and then determine the product of the target rule node as the target recommended product for the corresponding customer; or

[0113] If the first target rule node of the target layer is a leaf node, the product of the first target rule node is determined to be the target recommended product for the corresponding customer.

[0114] In this embodiment, the target recommended product may include one or more.

[0115] In this embodiment, a decision tree algorithm is used for each customer's behavior score, credit value, customer portrait and product portrait of all types of recommended products. The behavior score, credit value and customer portrait of each customer are matched layer by layer in the decision tree to determine the target recommended product, avoiding directly obtaining the recommended target recommended product based on a single indicator value of the customer. The target recommended product is recommended from multiple dimensions at the same time, thereby improving the recommendation accuracy of the target recommended product.

[0116] In this embodiment, the attribute node refers to the attribute node corresponding to multiple attributes obtained by combining the behavior score, credit value, and customer profile of each customer.

[0117] Furthermore, the decision tree is constructed based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products, including:

[0118] Determine attribute nodes based on each customer's behavior score, credit score, and customer profile;

[0119] Traverse all attribute nodes and product portraits of all recommended products of the aforementioned categories to obtain all recommended products of the categories corresponding to each attribute node as rule nodes;

[0120] The attribute with the largest weight is used as the root node in the attribute nodes;

[0121] If the rule node includes a category of recommended products, determining that the rule node is a leaf node;

[0122] If the rule node includes at least two categories of recommended products, determining that the rule node is a subtree;

[0123] The attribute with the largest weight except the upper layer is selected as the attribute node of the next layer of the subtree.

[0124] For example, the customer's behavior score is 9 points, the credit score is 9.5 points, and the customer profile is index fund-long-term holding. The order of the customer's attribute nodes is determined according to the weight value as follows: behavior score node-credit score node-index fund node-long-term holding node.

[0125] Product portraits of all types of recommended products: Product portraits of Category A recommended products: behavioral score greater than or equal to 9 points - credit value greater than or equal to 9 points - A01 stock fund - recommended score 9.2 points; Product portraits of Category B recommended products: behavioral score greater than or equal to 9 points - credit value greater than or equal to 9 points - A02 index fund - long-term holding - recommended score 9.5 points; Product portraits of Category C recommended products: behavioral score greater than or equal to 9 points - credit value range of 8 to 8.9 points - long-term holding - recommended score 9.0 points.

[0126] Traversing all the attribute nodes and the product portraits of all types of recommended products, the rule nodes corresponding to each attribute node of the customer are: the rule nodes corresponding to the behavior score of 9 points are: Class A recommended products, Class B recommended products and Class C products; the rule nodes corresponding to the credit value node of 9.5 points are: Class A recommended products and Class B recommended products; the rule node corresponding to the index fund node is: Class B recommended products; the rule node corresponding to the long-term holding node is: Class B recommended products.

[0127] The customer's behavior score of 9 points - credit value of 9.5 - index fund - long-term holding are input into the system as the customer's attributes. According to the decision tree and the customer's attributes: behavior score of 9 points, the attribute node behavior score node of the first layer of the decision tree is located, and the behavior score node is matched to obtain Class A recommended products, Class B recommended products and Class C products. It is determined that the Class A recommended products, Class B recommended products and Class C products are subtrees. Then, the system continues to locate the attribute node credit value node of the second layer of the decision tree below the Class A recommended products and Class B recommended products. The credit value node is matched to obtain Class A recommended products and Class B recommended products. It is determined that the Class A recommended products and Class B recommended products are subtrees. Then, the system continues to locate the attribute node index fund node of the third layer of the decision tree below the Class A recommended products and Class B recommended products. The index fund node is matched to the Class B recommended product. It is determined that the Class B recommended product is a leaf node, and the Class B recommended product is determined as the recommended target recommended product.

[0128] In this embodiment, since the decision tree has a clear hierarchy, the decision tree is constructed through each customer's behavior score, credit value, customer portrait and product portraits of all types of recommended products, which avoids the problem of missing attribute nodes in the process of determining the target recommended products. After the decision tree is constructed, each rule node in the decision tree is traversed to determine the target recommended product, thereby improving the accuracy of the determined target recommended products.

[0129] Furthermore, determining a target recommendation plan for each customer based on each customer's transaction willingness and the recommended product for the corresponding customer includes:

[0130] Compare each customer's transaction willingness with a preset transaction willingness threshold;

[0131] When each customer's transaction willingness is greater than or equal to the preset transaction willingness threshold, identifying the rate of return of each customer's recommended product, generating a recommendation plan based on the product information of multiple recommended products with higher rate of return, and determining the recommendation plan as the target recommendation plan for the corresponding customer; or

[0132] When the transaction willingness of each customer is less than the preset transaction willingness threshold, the target recommendation plan is rejected for the corresponding customer.

[0133] In this embodiment, multiple indicators of each user are calculated by using a pre-trained multiple indicator calculation model. Based on the indicator values ​​of each user's multiple indicators and the product portraits of all types of recommended products, a target recommendation plan for each customer is determined. Specifically, the target recommendation plan includes recommended target recommended products, product information related to the target recommended products, and a comparison analysis report with other products with high correlation.

[0134] In this embodiment, when determining the target recommendation plan, multiple indicators of each customer are taken into consideration, thereby improving the accuracy of the target recommendation plan.

[0135] For example, if the customer currently needs to use cash, make purchases in installments, make bill installments, increase or decrease the credit limit, make a temporary credit limit adjustment, apply for a large loan, etc., the recommended products will be those with high customer transaction willingness and high bank returns.

[0136] In other optional embodiments, after obtaining the target recommendation plan, the entire target recommendation plan process is recorded, and multiple feature values ​​are extracted from the recorded target plan recommendation process. The extracted multiple feature values ​​are divided based on multiple pre-trained indicator calculation models, and the divided feature sets are respectively input into the corresponding indicator calculation models for model training, thereby improving the calculation accuracy of the indicator calculation model.

[0137] In this embodiment, if the preset cycle is daily, the data of the previous day is collected in real time from multiple data sources, and the data collected from the previous day is cleaned, deduplicated and format converted using a big data processing method to ensure the accuracy of the user's subsequent target recommendation plan. At the same time, multiple indicator calculation models are trained every day to improve the accuracy of the indicator calculation model. Multiple indicator values ​​for each customer are calculated using multiple indicator calculation models that are updated in real time to ensure the accuracy of the calculated multiple indicator values. The target recommendation plan for each customer is determined in real time using the multiple indicator values ​​of each customer and the product portraits of all types of products. When confirming the target recommendation plan, multiple dimensions of the customer's card transactions or card usage scenarios are taken into consideration to generate a more accurate target recommendation plan, thereby improving the recommendation efficiency of the target recommendation plan and ensuring that the customer receives the optimal target recommendation plan in a short time, thereby improving customer satisfaction.

[0138] In summary, the intelligent recommendation method based on artificial intelligence described in this embodiment collects a first data set from multiple data sources according to a preset period, limits the data range of the first data set, and ensures the validity of the data collected in the first data set. The collected first data is pre-processed and cut to avoid the interference of errors and duplicate data, solves the problem of low efficiency and accuracy of subsequent data processing caused by inconsistent formats, improves the efficiency of product recommendations, and at the same time, the obtained target data set is distributed and operated, which improves the operating efficiency of the server and thus improves the efficiency of obtaining the target data set. Each of the target sub-datasets is input into a plurality of pre-trained indicator calculation models, and the multiple indicator values ​​of each customer are calculated. The target recommendation plan for each customer is determined based on the multiple indicator values ​​of each customer and the product portraits of all types of recommended products. When determining the target recommendation plan for each customer, the multiple indicators of each customer are taken into consideration, which improves the accuracy of the target recommendation plan.

[0139] Example 2

[0140] Figure 2 This is a structural diagram of an artificial intelligence-based intelligent recommendation device provided in Example 2 of the present invention.

[0141] In some embodiments, the artificial intelligence-based intelligent recommendation device 20 may include multiple functional modules composed of program code segments. The program code of each program segment in the artificial intelligence-based intelligent recommendation device 20 may be stored in the memory of the electronic device and executed by the at least one processor to perform (see Figure 1 Description) The function of intelligent recommendation based on artificial intelligence.

[0142] In this embodiment, the AI-based intelligent recommendation device 20 can be divided into multiple functional modules based on their functions. These modules may include a parsing module 201, a collection module 202, a segmentation module 203, an input module 204, and a determination module 205. As used herein, a module refers to a series of computer-readable instruction segments that can be executed by at least one processor and perform a fixed function, stored in a memory. The functions of each module in this embodiment will be described in detail in subsequent embodiments.

[0143] The parsing module 201 is used to parse the received intelligent recommendation request and obtain multiple data sources.

[0144] In this embodiment, in the financial field, due to the wide variety of products, customers cannot quickly determine the products they want to buy and need the server to recommend products. Therefore, the client initiates an intelligent recommendation request to the server. Specifically, the client can be a smart phone, IPAD or other existing smart devices, and the server can be an intelligent recommendation subsystem. During the product recommendation process, the client can send an intelligent recommendation request to the intelligent recommendation subsystem, and the intelligent recommendation subsystem is used to receive the intelligent recommendation request sent by the client.

[0145] In this embodiment, when the intelligent recommendation subsystem receives an intelligent recommendation request, it parses the message information in the intelligent recommendation request, and obtains multiple data sources through the calling interface according to the calling interface in the message information, wherein different data sources correspond to different calling interfaces. Specifically, the data sources may include: financial institution platforms, third-party platforms that have an association with the recommended products of the financial institution platforms, wherein the third-party platforms may include third-party payment platforms, bank transaction platforms, shopping platforms, third-party consumption platforms, etc.

[0146] Specifically, the intelligent recommendation request includes an intelligent recommendation task and a customer group corresponding to the intelligent recommendation task.

[0147] The collection module 202 is configured to collect a first data set corresponding to a period previous to the current period from the multiple data sources at every preset period, and pre-process each data in the first data set to obtain a second data set.

[0148] In this embodiment, considering the variability of data in the financial field, a data collection period can be preset. For example, the preset period can be set to 24 hours. If the current collection time is 9 o'clock, the data set within 24 hours before 9 o'clock is collected.

[0149] Specifically, the variability of data in the financial field refers to the rapid changes in customer data. For example, on day N, the customer's credit score is 9.5 points. On day N+1, the customer's credit score may become 2 points because the customer did not pay off the credit card in time or owed debts.

[0150] In an optional embodiment, the acquisition module 202 preprocesses each data in the first data set to obtain the second data set including:

[0151] Performing data cleaning and data deduplication on each data in the first data set to obtain a preprocessed first data set;

[0152] The format of each data in the preprocessed first data set is converted according to a preset format conversion rule, and each data after the format conversion is merged to obtain a second data set.

[0153] Specifically, the data cleaning includes one or more of the following methods: cleaning of missing values, format content, logical errors, and non-required data; and data deduplication refers to removing identical data.

[0154] Specifically, data conversion rules may be preset, for example, converting the data format of all date data into XXXX / XX / XX.

[0155] In this embodiment, the first data set is collected from multiple data sources according to a preset period, and the data range of the first data set is restricted, thereby ensuring the validity of the collected data of the first data set.

[0156] In this embodiment, by performing data cleaning, data deduplication and format conversion on each data in the first data set, the interference of erroneous and duplicate data is avoided, and at the same time, the problem of low efficiency and accuracy of subsequent data processing caused by inconsistent formats is solved, thereby improving the product recommendation efficiency.

[0157] The cutting module 203 is configured to cut the second data set to obtain a target data set.

[0158] In this embodiment, there is a large amount of financial data. In order to ensure the efficiency of subsequent data recommendation and improve the operating efficiency of the server, the second data set is segmented and the obtained target data set is distributed.

[0159] In an optional embodiment, the cutting module 203 cuts the second data set to obtain a target data set including:

[0160] The second data set is segmented according to a preset segmentation rule to obtain a target data set.

[0161] In other optional embodiments, the step of segmenting the second data set according to a preset segmentation rule to obtain a target data set includes:

[0162] Classifying the second data set according to the intelligent recommendation task in the intelligent recommendation request, and segmenting the classified second data set according to a preset first segmentation rule to obtain a second sub-data set;

[0163] Identifying the customer numbers of all customers in each of the second sub-datasets, and segmenting each of the second sub-datasets in the order of the last digits of the customer numbers of all customers according to a preset second segmentation rule to obtain a third sub-dataset corresponding to each of the second sub-datasets;

[0164] The third sub-dataset is merged, and the merged third sub-dataset is determined as the target dataset.

[0165] In this embodiment, the preset first cutting rule is set based on the task volume of the intelligent recommendation task in the intelligent recommendation request, and the preset second cutting rule is set based on the customer volume of all customers in the intelligent recommendation task.

[0166] For example, if the smart recommendation request contains 800,000 smart recommendation tasks, the 800,000 smart recommendation tasks are divided into 8 second sub-data sets according to the preset first cutting rule, and each second data set is cut according to the preset second cutting rule in the order of the last digits of the customer numbers of all customers to obtain the target data set corresponding to each second sub-data set, wherein the last digit refers to a serial number that is different from other customers.

[0167] In this embodiment, by segmenting the second data set and performing distributed operation on the obtained target data set, the operation efficiency of the server can be improved, thereby improving the efficiency of obtaining the target data set.

[0168] The input module 204 is configured to input each of the target sub-datasets in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer.

[0169] In this embodiment, the target data set includes multiple target sub-data sets. In order to better perform product recommendations, multiple indicator calculation models can be pre-trained, and multiple indicators of each customer in each of the target sub-data sets can be calculated separately through the preset multiple indicator calculation models.

[0170] In an optional embodiment, the input module 204 inputs each target sub-dataset in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer including:

[0171] Inputting each target sub-dataset into a pre-trained response rate calculation model to obtain the response rate of each customer;

[0172] Input each target sub-dataset into a pre-trained willingness calculation model to obtain the willingness of each customer;

[0173] Input each target sub-dataset into a pre-trained behavior score calculation model to obtain a behavior score for each customer;

[0174] Input each target sub-dataset into a pre-trained credit score calculation model to obtain the credit score of each customer;

[0175] Input each target sub-dataset into a pre-trained customer profile calculation model to obtain a customer profile for each customer;

[0176] The response rate, willingness, behavior score, credit value and customer portrait of each customer are determined as multiple indicator values ​​for each customer.

[0177] In this embodiment, the indicator calculation model may include other indicator calculation models such as a response rate calculation model, a willingness calculation model, a behavior scoring calculation model, a credit value calculation model and a customer portrait calculation model. The indicator calculation model is pre-trained according to the recommendation needs, wherein the training process of the multiple indicator calculation models is the same, the difference lies in the different input sample data sets. Specifically, for the training of the response rate calculation model, the input sample data set is: feature values ​​such as customer visits to products and purchase records; for the training of the willingness calculation model, the input sample data set is: feature values ​​such as customer scores and comments on each product; for the training of the behavior scoring calculation model, the input sample data set is: behavioral features such as average balance, number of overdue payments, and credit limit; for the training of the credit value calculation model, the input sample data set is: feature values ​​such as customer loan repayment records, credit information, and credit card usage records; for the training of the customer portrait calculation model, the input sample data set is: feature values ​​such as customer age, gender, and product preferences.

[0178] Specifically, the training process of the indicator calculation model includes: dividing a training set and a test set from the sample data set; inputting the training set into a preset neural network for training to obtain an indicator calculation model; inputting the test set into the indicator calculation model for testing, and calculating the test pass rate; if the test pass rate is greater than or equal to a preset pass rate threshold, determining that the training of the indicator calculation model is completed; if the test pass rate is less than the preset pass rate threshold, increasing the number of training sets and re-training the indicator calculation model.

[0179] The determination module 205 is used to determine a target recommendation plan for the corresponding customer based on the multiple indicator values ​​of each customer and the product portraits of all types of recommended products.

[0180] In this embodiment, due to the wide variety of product types in the financial industry, when determining the target recommendation plan, it is necessary to obtain the product information of all recommended products, and input the product information of all products into a pre-trained classification model to obtain the product labels of each type of recommended products, and construct a product portrait of each type of product based on the product labels of each type of recommended products, thereby improving the recommendation efficiency of subsequent target recommendation plans.

[0181] In an optional embodiment, the determination module 205 determines the target recommendation plan for the corresponding customer based on the multiple indicator values ​​of each customer and the product profiles of all recommended products, including:

[0182] Determine each customer's willingness to close a deal based on each customer's response rate and corresponding willingness;

[0183] Determine the target recommended products for each customer based on their behavior score, credit score, customer profile, and product profiles of all recommended products;

[0184] Based on each customer's willingness to make a deal and the target recommended products for the corresponding customer, determine the target recommendation plan for the corresponding customer.

[0185] In this embodiment, the response rate refers to the customer's response to the product, that is, whether the customer opens the product link, etc.; the willingness refers to the customer's willingness to the product; and the transaction willingness refers to the customer's willingness to complete the transaction.

[0186] Furthermore, determining each customer's transaction willingness based on each customer's response rate and corresponding willingness includes:

[0187] Calculate the product of each customer's response rate and a preset first weight value to obtain a first score;

[0188] Calculate the product of each customer's willingness and a preset second weight value to obtain a second score;

[0189] The sum of the first score and the second score is calculated to obtain each customer's transaction willingness.

[0190] In this embodiment, when calculating each customer's willingness to complete a transaction, taking into account the differences in the impact of response rate and willingness on the willingness to complete a transaction, a first weight value is set in advance for the response rate, and a second weight value is set in advance for the willingness. The first weight value and the second weight value can be set equal or unequal according to the recommended task situation, and this embodiment does not impose any restrictions on this.

[0191] Furthermore, the target recommended products for each customer are determined based on the behavior score, credit score, customer profile, and product profiles of all recommended products of each customer, including:

[0192] Construct a decision tree based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products. Each layer of the decision tree includes attribute nodes and rule nodes corresponding to the attribute nodes. The attribute nodes refer to attribute nodes corresponding to multiple attributes obtained by combining each customer's behavior score, credit score, and customer profile.

[0193] Locating a target attribute node corresponding to the attribute of each customer in the decision tree to obtain a first target rule node corresponding to the target attribute node in a target layer;

[0194] If the first target rule node of the target layer is a subtree, continue locating according to the attribute node of the next layer of the subtree to obtain a second target rule node corresponding to the attribute node of the next layer, until the target rule node is determined to be a leaf node, and then determine the product of the target rule node as the target recommended product for the corresponding customer; or

[0195] If the first target rule node of the target layer is a leaf node, the product of the first target rule node is determined to be the target recommended product for the corresponding customer.

[0196] In this embodiment, the target recommended product may include one or more.

[0197] In this embodiment, a decision tree algorithm is used for each customer's behavior score, credit value, customer portrait and product portrait of all types of recommended products. The behavior score, credit value and customer portrait of each customer are matched layer by layer in the decision tree to determine the target recommended product, avoiding directly obtaining the recommended target recommended product based on a single indicator value of the customer. The target recommended product is recommended from multiple dimensions at the same time, thereby improving the recommendation accuracy of the target recommended product.

[0198] In this embodiment, the attribute node refers to the attribute node corresponding to multiple attributes obtained by combining the behavior score, credit value, and customer profile of each customer.

[0199] Furthermore, the decision tree is constructed based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products, including:

[0200] Determine attribute nodes based on each customer's behavior score, credit score, and customer profile;

[0201] Traverse all attribute nodes and product portraits of all recommended products of the aforementioned categories to obtain all recommended products of the categories corresponding to each attribute node as rule nodes;

[0202] The attribute with the largest weight is used as the root node in the attribute nodes;

[0203] If the rule node includes a category of recommended products, determining that the rule node is a leaf node;

[0204] If the rule node includes at least two categories of recommended products, determining that the rule node is a subtree;

[0205] The attribute with the largest weight except the upper layer is selected as the attribute node of the next layer of the subtree.

[0206] For example, the customer's behavior score is 9 points, the credit score is 9.5 points, and the customer profile is index fund-long-term holding. The order of the customer's attribute nodes is determined according to the weight value as follows: behavior score node-credit score node-index fund node-long-term holding node.

[0207] Product portraits of all types of recommended products: Product portraits of Category A recommended products: behavioral score greater than or equal to 9 points - credit value greater than or equal to 9 points - A01 stock fund - recommended score 9.2 points; Product portraits of Category B recommended products: behavioral score greater than or equal to 9 points - credit value greater than or equal to 9 points - A02 index fund - long-term holding - recommended score 9.5 points; Product portraits of Category C recommended products: behavioral score greater than or equal to 9 points - credit value range of 8 to 8.9 points - long-term holding - recommended score 9.0 points.

[0208] Traversing all the attribute nodes and the product portraits of all types of recommended products, the rule nodes corresponding to each attribute node of the customer are: the rule nodes corresponding to the behavior score of 9 points are: Class A recommended products, Class B recommended products and Class C products; the rule nodes corresponding to the credit value node of 9.5 points are: Class A recommended products and Class B recommended products; the rule node corresponding to the index fund node is: Class B recommended products; the rule node corresponding to the long-term holding node is: Class B recommended products.

[0209] The customer's behavior score of 9 points - credit value of 9.5 - index fund - long-term holding are input into the system as the customer's attributes. According to the decision tree and the customer's attributes: behavior score of 9 points, the attribute node behavior score node of the first layer of the decision tree is located, and the behavior score node is matched to obtain Class A recommended products, Class B recommended products and Class C products. It is determined that the Class A recommended products, Class B recommended products and Class C products are subtrees. Then, the system continues to locate the attribute node credit value node of the second layer of the decision tree below the Class A recommended products and Class B recommended products. The credit value node is matched to obtain Class A recommended products and Class B recommended products. It is determined that the Class A recommended products and Class B recommended products are subtrees. Then, the system continues to locate the attribute node index fund node of the third layer of the decision tree below the Class A recommended products and Class B recommended products. The index fund node is matched to the Class B recommended product. It is determined that the Class B recommended product is a leaf node, and the Class B recommended product is determined as the recommended target recommended product.

[0210] In this embodiment, since the decision tree has a clear hierarchy, the decision tree is constructed through each customer's behavior score, credit value, customer portrait and product portraits of all types of recommended products, which avoids the problem of missing attribute nodes in the process of determining the target recommended products. After the decision tree is constructed, each rule node in the decision tree is traversed to determine the target recommended product, thereby improving the accuracy of the determined target recommended products.

[0211] Furthermore, determining a target recommendation plan for each customer based on each customer's transaction willingness and the recommended product for the corresponding customer includes:

[0212] Compare each customer's transaction willingness with a preset transaction willingness threshold;

[0213] When each customer's transaction willingness is greater than or equal to the preset transaction willingness threshold, identifying the rate of return of each customer's recommended product, generating a recommendation plan based on the product information of multiple recommended products with higher rate of return, and determining the recommendation plan as the target recommendation plan for the corresponding customer; or

[0214] When the transaction willingness of each customer is less than the preset transaction willingness threshold, the target recommendation plan is rejected for the corresponding customer.

[0215] In this embodiment, multiple indicators of each user are calculated by using a pre-trained multiple indicator calculation model. Based on the indicator values ​​of each user's multiple indicators and the product portraits of all types of recommended products, a target recommendation plan for each customer is determined. Specifically, the target recommendation plan includes recommended target recommended products, product information related to the target recommended products, and a comparison analysis report with other products with high correlation.

[0216] In this embodiment, when determining the target recommendation plan, multiple indicators of each customer are taken into consideration, thereby improving the accuracy of the target recommendation plan.

[0217] For example, if the customer currently needs to use cash, make purchases in installments, make bill installments, increase or decrease the credit limit, make a temporary credit limit adjustment, apply for a large loan, etc., the recommended products will be those with high customer transaction willingness and high bank returns.

[0218] In other optional embodiments, after obtaining the target recommendation plan, the entire target recommendation plan process is recorded, and multiple feature values ​​are extracted from the recorded target plan recommendation process. The extracted multiple feature values ​​are divided based on multiple pre-trained indicator calculation models, and the divided feature sets are respectively input into the corresponding indicator calculation models for model training, thereby improving the calculation accuracy of the indicator calculation model.

[0219] In this embodiment, if the preset cycle is daily, the data of the previous day is collected in real time from multiple data sources, and the data collected from the previous day is cleaned, deduplicated and format converted using a big data processing method to ensure the accuracy of the user's subsequent target recommendation plan. At the same time, multiple indicator calculation models are trained every day to improve the accuracy of the indicator calculation model. Multiple indicator values ​​for each customer are calculated using multiple indicator calculation models that are updated in real time to ensure the accuracy of the calculated multiple indicator values. The target recommendation plan for each customer is determined in real time using the multiple indicator values ​​of each customer and the product portraits of all types of products. When confirming the target recommendation plan, multiple dimensions of the customer's card transactions or card usage scenarios are taken into consideration to generate a more accurate target recommendation plan, thereby improving the recommendation efficiency of the target recommendation plan and ensuring that the customer receives the optimal target recommendation plan in a short time, thereby improving customer satisfaction.

[0220] In summary, the intelligent recommendation device based on artificial intelligence described in this embodiment collects a first data set from multiple data sources according to a preset period, limits the data range of the first data set, and ensures the validity of the data collected in the first data set. The collected first data is pre-processed and cut to avoid the interference of errors and duplicate data, solve the problem of low efficiency and accuracy of subsequent data processing caused by inconsistent formats, and improve the efficiency of product recommendations. At the same time, the obtained target data set is distributed and operated, which improves the operating efficiency of the server and thus improves the efficiency of obtaining the target data set. Each of the target sub-datasets is input into a plurality of pre-trained indicator calculation models, and the multiple indicator values ​​of each customer are calculated. According to the multiple indicator values ​​of each customer and the product portraits of all types of recommended products, the target recommendation plan for each customer is determined. When determining the target recommendation plan for each customer, the multiple indicators of each customer are taken into consideration, which improves the accuracy of the target recommendation plan.

[0221] Example 3

[0222] See Figure 3 FIG. 3 is a schematic diagram of the structure of an electronic device according to Embodiment 3 of the present invention. In a preferred embodiment of the present invention, the electronic device 3 includes a memory 31 , at least one processor 32 , at least one communication bus 33 and a transceiver 34 .

[0223] Those skilled in the art should understand that Figure 3 The structure of the electronic device shown does not constitute a limitation of the embodiments of the present invention, and can be either a bus structure or a star structure. The electronic device 3 can also include more or less other hardware or software than shown in the figure, or a different component arrangement.

[0224] In some embodiments, the electronic device 3 is an electronic device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits, programmable gate arrays, digital processors, and embedded devices. The electronic device 3 may also include a client device, which includes but is not limited to any electronic product capable of human-computer interaction with a client via a keyboard, mouse, remote control, touchpad, or voice-controlled device, such as a personal computer, tablet computer, smartphone, digital camera, etc.

[0225] It should be noted that the electronic device 3 is only an example. Other existing or future electronic products that are suitable for the present invention should also be included in the protection scope of the present invention and included here by reference.

[0226] In some embodiments, the memory 31 is used to store program code and various data, such as the artificial intelligence-based intelligent recommendation device 20 installed in the electronic device 3, and to achieve high-speed and automatic access to programs or data during the operation of the electronic device 3. The memory 31 includes read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electronically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0227] In some embodiments, the at least one processor 32 may be composed of an integrated circuit, such as a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The at least one processor 32 is the control core (Control Unit) of the electronic device 3, connecting the various components of the entire electronic device 3 using various interfaces and lines, and executing or executing programs or modules stored in the memory 31, as well as calling data stored in the memory 31, to perform various functions of the electronic device 3 and process data.

[0228] In some embodiments, the at least one communication bus 33 is configured to implement connection and communication between the memory 31 and the at least one processor 32 .

[0229] Although not shown, the electronic device 3 may also include a power source (such as a battery) to power various components. Optionally, the power source may be logically connected to the at least one processor 32 via a power management device, thereby enabling the power management device to manage charging, discharging, and power consumption. The power source may also include one or more DC or AC power sources, a recharging device, a power failure detection circuit, a power converter or inverter, a power status indicator, and other arbitrary components. The electronic device 3 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be detailed here.

[0230] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0231] The above-mentioned integrated unit implemented in the form of a software functional module can be stored in a computer-readable storage medium. The above-mentioned software functional module stored in a storage medium includes a number of instructions for causing a computer device (which can be a personal computer, electronic device, or network device, etc.) or a processor to execute portions of the methods described in various embodiments of the present invention.

[0232] In a further embodiment, in combination Figure 2 The at least one processor 32 can execute the operating device of the electronic device 3 and various installed applications (such as the artificial intelligence-based intelligent recommendation device 20), program codes, etc., for example, the above-mentioned modules.

[0233] The memory 31 stores program codes, and the at least one processor 32 can call the program codes stored in the memory 31 to execute related functions. For example, Figure 2 The modules described are program codes stored in the memory 31 and executed by the at least one processor 32, thereby realizing the functions of the modules to achieve the purpose of intelligent recommendation based on artificial intelligence.

[0234] Exemplarily, the program code may be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 32 to complete the present application. The one or more modules / units may be a series of computer-readable instruction segments capable of performing specific functions, which are used to describe the execution process of the program code in the electronic device 3. For example, the program code may be divided into a parsing module 201, a collection module 202, a cutting module 203, an input module 204, and a determination module 205.

[0235] In one embodiment of the present invention, the memory 31 stores a plurality of computer-readable instructions, and the plurality of computer-readable instructions are executed by the at least one processor 32 to implement a function of intelligent recommendation based on artificial intelligence.

[0236] Specifically, the specific implementation method of the at least one processor 32 for the above instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0237] In the several embodiments provided by the present invention, 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 illustrative. For example, the module division is merely a logical function division, and actual implementation may employ other division methods.

[0238] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, and may be located in one place or distributed across multiple network elements. Some or all of the modules may be selected to achieve the purpose of this embodiment based on actual needs.

[0239] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing unit, 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 hardware plus software functional modules.

[0240] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be regarded as exemplary and non-restrictive from any point of view, and the scope of the invention is defined by the appended claims rather than the above description, and it is intended that all variations that fall within the meaning and scope of the equivalents of the claims be encompassed within the present invention. Any reference numerals in the claims should not be construed as limiting the claims to which they relate. Furthermore, it is apparent that the word "comprising" does not exclude other units or, and the singular does not exclude the plural. The multiple units or devices set forth in the present invention may also be implemented by one unit or device through software or hardware. Words such as first and second are used to indicate names and do not indicate any particular order.

[0241] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent recommendation method based on artificial intelligence, characterized in that: The method comprises: Parse the received intelligent recommendation request and obtain multiple data sources; Collecting a first data set corresponding to a previous period of a current period from the multiple data sources at every preset period, and preprocessing each data in the first data set to obtain a second data set; Cutting the second data set to obtain a target data set includes: classifying the second data set according to the intelligent recommendation task in the intelligent recommendation request, and cutting the classified second data set according to a preset first cutting rule to obtain a second sub-data set, wherein the preset first cutting rule is set based on the task amount of the intelligent recommendation task in the intelligent recommendation request; identifying the customer numbers of all customers in each of the second sub-data sets, and cutting each of the second sub-data sets in the order of the last digits of the customer numbers of all customers according to the preset second cutting rule to obtain a third sub-data set corresponding to each of the second sub-data sets, wherein the preset second cutting rule is set based on the customer amount of all customers in the intelligent recommendation task; merging the third sub-data sets, and determining the merged third sub-data set as the target data set, wherein the target data set contains multiple target sub-data; Inputting each target sub-dataset in the target data set into a pre-trained multiple indicator calculation model to obtain multiple indicator values ​​for each customer, the multiple indicator values ​​including the response rate, willingness, behavior score, credit value and customer profile of each customer; Based on the multiple indicator values ​​of each customer and the product portraits of all types of recommended products, a target recommendation plan for the corresponding customer is determined, including: determining each customer's transaction willingness based on each customer's response rate and corresponding willingness; determining each customer's target recommended product based on each customer's behavior score, credit value, customer portrait and product portraits of all types of recommended products; determining the target recommendation plan for the corresponding customer based on each customer's transaction willingness and the corresponding customer's target recommended product.

2. The artificial intelligence-based intelligent recommendation method according to claim 1, wherein: Inputting each target sub-dataset in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer includes: Inputting each target sub-dataset into a pre-trained response rate calculation model to obtain the response rate of each customer; Input each target sub-dataset into a pre-trained willingness calculation model to obtain the willingness of each customer; Input each target sub-dataset into a pre-trained behavior score calculation model to obtain a behavior score for each customer; Input each target sub-dataset into a pre-trained credit score calculation model to obtain the credit score of each customer; Input each target sub-dataset into a pre-trained customer profile calculation model to obtain a customer profile for each customer; The response rate, willingness, behavior score, credit value and customer portrait of each customer are determined as multiple indicator values ​​for each customer.

3. The artificial intelligence-based intelligent recommendation method according to claim 2, wherein: Determining each customer's transaction willingness based on each customer's response rate and corresponding willingness includes: Calculate the product of each customer's response rate and a preset first weight value to obtain a first score; Calculate the product of each customer's willingness and a preset second weight value to obtain a second score; The sum of the first score and the second score is calculated to obtain each customer's transaction willingness.

4. The artificial intelligence-based intelligent recommendation method according to claim 3, wherein: The target recommended products for each customer are determined based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products, including: Construct a decision tree based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products. Each layer of the decision tree includes attribute nodes and rule nodes corresponding to the attribute nodes. The attribute nodes refer to attribute nodes corresponding to multiple attributes obtained by combining each customer's behavior score, credit score, and customer profile. Locating a target attribute node corresponding to the attribute of each customer in the decision tree to obtain a first target rule node corresponding to the target attribute node in a target layer; If the first target rule node of the target layer is a subtree, continue locating according to the attribute node of the next layer of the subtree to obtain a second target rule node corresponding to the attribute node of the next layer, until the target rule node is determined to be a leaf node, and then determine the product of the target rule node as the target recommended product for the corresponding customer; or If the first target rule node of the target layer is a leaf node, the product of the first target rule node is determined to be the target recommended product for the corresponding customer.

5. The artificial intelligence-based intelligent recommendation method according to claim 4, characterized in that: The decision tree constructed based on each customer's behavior score, credit score, customer profile, and product profiles of all recommended products includes: Determine attribute nodes based on each customer's behavior score, credit score, and customer profile; Traverse all attribute nodes and product portraits of all recommended products of the aforementioned categories to obtain all recommended products of the categories corresponding to each attribute node as rule nodes; The attribute with the largest weight is used as the root node in the attribute nodes; If the rule node includes a category of recommended products, determining that the rule node is a leaf node; If the rule node includes at least two categories of recommended products, determining that the rule node is a subtree; The attribute with the largest weight except the upper layer is selected as the attribute node of the next layer of the subtree.

6. An intelligent recommendation device based on artificial intelligence, characterized in that: The device comprises: The parsing module is used to parse the received intelligent recommendation request and obtain multiple data sources; an acquisition module, configured to acquire, from the plurality of data sources at predetermined intervals, a first data set corresponding to a period preceding a current period, and preprocess each data in the first data set to obtain a second data set; The cutting module is used to cut the second data set to obtain a target data set, including: classifying the second data set according to the intelligent recommendation task in the intelligent recommendation request, and cutting the classified second data set according to a preset first cutting rule to obtain a second sub-data set, wherein the preset first cutting rule is set based on the task amount of the intelligent recommendation task in the intelligent recommendation request; identifying the customer numbers of all customers in each of the second sub-data sets, and cutting each of the second sub-data sets in the order of the last digits of the customer numbers of all customers according to the preset second cutting rule to obtain a third sub-data set corresponding to each of the second sub-data sets, wherein the preset second cutting rule is set based on the customer amount of all customers in the intelligent recommendation task; merging the third sub-data sets, and determining the merged third sub-data set as the target data set, wherein the target data set contains multiple target sub-data; An input module is configured to input each target sub-dataset in the target data set into a plurality of pre-trained indicator calculation models to obtain a plurality of indicator values ​​for each customer, wherein the plurality of indicator values ​​include a response rate, willingness, behavior score, credit score, and customer profile of each customer; A determination module is used to determine a target recommendation plan for the corresponding customer based on multiple indicator values ​​of each customer and product portraits of all types of recommended products, including: determining each customer's transaction willingness based on each customer's response rate and corresponding willingness; determining each customer's target recommended product based on each customer's behavior score, credit value, customer portrait and product portraits of all types of recommended products; determining the target recommendation plan for the corresponding customer based on each customer's transaction willingness and the corresponding customer's target recommended product.

7. An electronic device, characterized in that: The electronic device includes a processor and a memory, and the processor is used to implement the artificial intelligence-based intelligent recommendation method according to any one of claims 1 to 5 when executing the computer program stored in the memory.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the artificial intelligence-based intelligent recommendation method according to any one of claims 1 to 5 is implemented.

Images