A product recommendation method, apparatus, device, and medium

Abstract

Embodiments of the present application disclose a product recommendation method, device, equipment and medium. The method comprises: forming a product pool from the dimensions of customer attribute similarity, financial attribute similarity and event marketing; if the target customer is a customer without interaction history, inputting customer attribute information and product attribute information into a first recommendation model to obtain a first probability of purchasing a product in the product pool, and recommending the product to the customer based on the first probability; and if the target customer is a customer with interaction history, inputting customer attribute information, product attribute information and product interaction attribute information into a second recommendation model to obtain a second probability of purchasing a product in the product pool, and recommending the product to the customer based on the second probability. The embodiments can solve the problem of data skewing caused by too detailed product force leading to fewer customers purchasing the product, and can also solve the cold start problem caused by lack of customer information.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a product recommendation method, apparatus, device, and storage medium. Background Technology

[0002] In the financial sector, diversified product recommendations that cover all customer groups have become an important means to enhance customer loyalty, improve product marketing efficiency, promote capital accumulation, and optimize customer asset allocation.

[0003] In related technologies, product recommendations rely more on customers' historical holdings and purchase information, lacking a cold start strategy; and without forming a suitable product pool, it is easy to result in overly detailed product offerings, leading to fewer customers purchasing the products and creating data bias. Summary of the Invention

[0004] This invention provides a product recommendation method, apparatus, device, and storage medium to address the problem of data skew caused by too much product detail resulting in fewer customers purchasing the product, and to solve the cold start problem caused by a lack of customer information.

[0005] According to one aspect of the present invention, a product recommendation method is provided, comprising:

[0006] A product pool is formed based on customer attribute similarity, financial attribute similarity, and event marketing.

[0007] If the target customer is a customer with no interaction history, the customer attribute information of the target customer and the product attribute information of each product in the product pool are input into the first recommendation model to obtain the probability of purchasing a product in the product pool, which is used as the first probability, and product recommendations are made to the target customer based on the first probability.

[0008] If the target customer is a customer with an interaction history, the customer attribute information of the target customer, the product attribute information of each product in the product pool, and the product interaction attribute information are input into the second recommendation model to obtain the probability of purchasing a product in the product pool, which is then used as the second probability, and product recommendations are made to the target customer based on the second probability.

[0009] According to another aspect of the present invention, a product recommendation device is provided, comprising:

[0010] The product pool formation module is used to form a product pool based on customer attribute similarity, financial attribute similarity, and event marketing.

[0011] The first recommendation module is used to input the customer attribute information of the target customer and the product attribute information of each product in the product pool into the first recommendation model if the target customer is a customer with no interaction history, to obtain the probability of purchasing a product in the product pool, and use it as the first probability, and to recommend products to the target customer based on the first probability.

[0012] The second recommendation module is used to input the customer attribute information of the target customer, the product attribute information of each product in the product pool, and the product interaction attribute information into the second recommendation model if the target customer is a customer with an interaction history, to obtain the probability of purchasing a product in the product pool, and use it as the second probability, and to recommend products to the target customer based on the second probability.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] At least one processor; and

[0015] A memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform a product recommendation method according to any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement a product recommendation method according to any embodiment of the present invention.

[0018] The technical solution of this invention addresses the problem of data skew caused by a small number of customers purchasing products due to overly detailed product offerings. On the one hand, for customers without interaction history, customer attribute information and product attribute information can be input into the model, and the probability of the customer purchasing products can be obtained based on the model. This can solve the cold start problem caused by a lack of customer information by recommending products to customers.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a product recommendation method provided according to Embodiment 1 of the present invention;

[0022] Figure 2a This is a flowchart of another product recommendation method provided according to Embodiment 2 of the present invention;

[0023] Figure 2b This is a wide table provided according to Embodiment 2 of the present invention for customers without an interaction history;

[0024] Figure 2c This is a wide table provided according to Embodiment 2 of the present invention for customers with an interaction history;

[0025] Figure 3 This is a flowchart of a product recommendation device provided according to Embodiment 4 of the present invention;

[0026] Figure 4 This is a schematic diagram of the structure of an electronic device that implements a product recommendation method according to an embodiment of the present invention. Detailed Implementation

[0027] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] Example 1

[0030] Figure 1 This is a flowchart illustrating a product recommendation method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where product recommendations are made to customers. The method can be executed by a product recommendation device, which can be implemented in hardware and / or software and can be integrated into an electronic device. Figure 1 As shown, the method includes:

[0031] S110. Form a product pool based on customer attribute similarity, financial attribute similarity, and event marketing dimensions.

[0032] The product pool can be formed by three dimensions: customer attribute similarity, financial attribute similarity, and event marketing. Based on this product pool, suitable products are recommended to each user. In this embodiment, financial products can be recommended to customers based on the product pool. This embodiment does not impose specific restrictions on this.

[0033] In this embodiment, customer attribute similarity can be understood as statistically analyzing the correlation between customer-related information and products. For example, it can distinguish between customer groups that have formed a product pool and those that have not. If the proportion of customers who have not formed a product pool is 80%, it means that these customers do not hold the corresponding bank's wealth management, deposit, insurance, and fund products. Therefore, it is necessary to construct the correlation between these customers and the products through other relevant information. For example, it can determine the customer attribute similarity based on customer information. Specifically, it can form a specific customer attribute similarity based on the typical information of customers such as payroll customers, mid-to-high-end customers, and credit card customers.

[0034] In this embodiment, financial attribute similarity can be understood as statistically analyzing a customer's historical holding or purchasing behavior of products, and constructing a collaborative filtering model to increase the probability of a product being recommended.

[0035] In this embodiment, event marketing can be understood as an event-driven stimulus, such as fund preparation or product maturity, that typically precedes a customer's purchase of a certain type of asset management product. Therefore, it is necessary to construct association rules between the customer's current account changes, transaction history, and product purchase and redemption behaviors to identify events strongly correlated with product purchases.

[0036] For example, S110 may include: in the customer attribute similarity dimension, selecting a first variable that has a significant impact on customer product purchases and is free from collinearity based on customer group information, basic attribute information, and account management attribute information through a decision tree; performing equal-frequency binning on the first variable and encoding the binned first variable to obtain first encoded information; inputting the first encoded information into a locality-sensitive hash model to obtain the customer's first feature vector; and placing products purchased by the customer within a preset time period corresponding to feature vectors similar to the first feature vector into a product pool; in the financial attribute similarity dimension, selecting a second variable that has a significant impact on customer product purchases and is free from collinearity based on customer product purchase information, product holding information, and financial asset information through a decision tree; performing equal-frequency binning on the second variable and encoding the binned second variable to obtain second encoded information; inputting the second encoded information into a locality-sensitive hash model to obtain the customer's second feature vector; and placing products purchased by the customer within a preset time period corresponding to feature vectors similar to the second feature vector into a product pool; in the event marketing dimension, if a customer experiences a target event, placing products corresponding to the target event into the product pool.

[0037] In this embodiment, the first feature vector and the second feature vector of a customer can be determined on the dimensions of customer attribute similarity and financial attribute similarity, respectively, and the products purchased by customers whose feature vectors are similar to the first feature vector within a preset time period can be placed into the product pool.

[0038] For the customer attribute similarity dimension, we can start from customer group information (payroll customers, mid-to-high-end customers, and credit card customers, etc.), basic attributes, account management attributes, etc., and use decision trees to filter out the first variable that has a greater impact on customers' product purchases and is not collinear.

[0039] Regarding the similarity of financial attributes, a second variable that has a significant impact on customer product purchases and is not collinear can be selected through decision trees based on customer product purchase information, product holding information, and financial asset information.

[0040] In this embodiment, equal-frequency binning can be performed on the first variable and the second variable respectively. By equal-frequency binning, each customer can be tagged according to the amount disbursed and the tags can be encoded to obtain the first and second encoded information. The first and second encoded information are then input into the Locality Sensitive Hash Model to obtain the first and second feature vectors of the customer. It should be emphasized that this embodiment does not specifically limit the type of tagging and can be flexibly set as needed.

[0041] Specifically, in this embodiment, for the equal-frequency binning of the first variable, taking 10 bins as an example, customers whose actual payout amount is at the 58th percentile are labeled as "actual payout amount greater than 40% and less than 60%" and "actual payout amount greater than 50% and less than 70%". In this way, customers with similar payout amounts in the same bin will share two labels, while customers with similar payout amounts in different bins will share one label. Based on this labeling system, each customer will be assigned multiple labels (binary variables such as gender will only generate one label). Customers who have made purchases within the past 3 months of the customer attribute data collection period are selected, and their labels are first encoded to construct a locality-sensitive hashing model, obtaining the first feature vector. After labeling and encoding the predicted customers, the model can query the most similar customers (2 people) and recommend products purchased by these similar customers in the past 3 months of the customer attribute data collection period.

[0042] Specifically, in this embodiment, for the equal-frequency binning of the second variable, taking 10 bins as an example, customers whose actual payout amount is at the 58th percentile are labeled as "actual payout amount greater than 40% and less than 60%" and "actual payout amount greater than 50% and less than 70%". In this way, customers with similar payout amounts in the same bin will share two labels, while customers with similar payout amounts in different bins will share one label. Based on this labeling system, each customer will be assigned multiple labels (binary variables such as gender will only generate one label). Customers who have made purchases within the past 3 months of the statistical period are selected, and their labels are second-encoded to construct a locality-sensitive hashing model to obtain a second feature vector. After labeling and encoding the predicted customers, the model can query the most similar customers (5 people) and recommend products purchased by these similar customers in the past 3 months of the customer attribute data statistical period.

[0043] Optionally, the encoding method used in this embodiment may be Minhash encoding, and this embodiment does not specifically limit it.

[0044] For example, adding products purchased by customers corresponding to feature vectors similar to the second feature vector within a preset time period to the product pool may include: if a customer corresponding to a feature vector similar to the second feature vector purchased a first type of product within the preset time period, and the similarity between the first type of product and the second type of product is higher than a preset threshold, then the first type of product and the second type of product are added to the product pool; if a customer corresponding to a feature vector similar to the second feature vector purchased a first type of product within the preset time period, and the number of first type of products is less than a first preset quantity, then all products in the category to which the first type of product belongs are added to the product pool; if a customer corresponding to a feature vector similar to the second feature vector purchased a target product within the preset time period, and the number of target products purchased is greater than a second preset quantity, then, under the target product category, the products ranked in the top n by number of purchases are added to the product pool.

[0045] In this embodiment, if the product recommendation granularity is too fine for customers with historical holdings or purchases, resulting in a small number of purchasers for a single product category and consequently a small number of purchasers for both product categories, data skew will occur when calculating product similarity. Consequently, the association rules cannot output rules that strongly support business interpretation. Therefore, products are aggregated based on terms, interest rates, and product attributes. Starting from the aggregated products held by the customer, the purchase status is used as a collaborative filtering score, and the similarity between products is calculated using the Jaccard similarity coefficient. For example, if a customer has purchased product A (category) within the past 3 months and the similarity between product A (category) and product B (category) is higher than the threshold (0.1), then product B (category) can be recommended to them. When the number of products (category) is small, all products under that category are recommended based on the previously analyzed product aggregate. When the number of products (category) is large, the top 3 product categories with the most purchasers in the previous month are recommended. When there is a business promotion need, a business promotion product series is recommended. In addition to cross-product similarity, products with a high probability of repeated purchases within three months should also be included in the category of highly similar products; otherwise, they should not be included in the category of AA similar combinations.

[0046] Specifically, Table 1 below shows the repeat purchases of product A and product B by different customers:

[0047] Table 1

[0048] Customer 1 Customer 2 Customer 3 Customer 4 Product A Buy Buy Buy Product B Buy Buy Buy

[0049] Furthermore, taking Table 1 above as an example, the Jaccard similarity coefficient can be calculated using the following formula, and the result can be 2 / 4 = 5:

[0050]

[0051] For example, the target event may include product purchase-related events constructed from current account change information, transaction history information, and product purchase and redemption information.

[0052] In this embodiment, in terms of event marketing, if it is determined that a customer has experienced a target event, namely, an event related to product purchase constructed from current account change information, transaction flow information, and product purchase and redemption information, then the product corresponding to the target event will be placed into the product pool.

[0053] In this embodiment, the target events may further include business experience and business needs. Based on one or more target events that together construct a product purchase-related event, when a customer experiences one of these target events, the system can recommend corresponding products to the customer based on the target events.

[0054] S120. If the target customer is a customer without an interaction history, input the target customer's customer attribute information and the product attribute information of each product in the product pool into the first recommendation model to obtain the probability of purchasing a product in the product pool, and use it as the first probability, and make product recommendations to the target customer based on the first probability.

[0055] Interaction history can be understood as the customer's past transactions with the corresponding bank, including but not limited to purchasing wealth management products and trading wealth management products.

[0056] For example, customer attribute information may include a first variable; product attribute information may include product sub-category information, product category information, product term information, product liquidity information, and product profitability information.

[0057] Among them, the first variable based on customer attribute information can be a variable that has a significant impact on customers' purchase of asset management products, obtained from the customer attribute similarity dimension.

[0058] In this embodiment of the invention, products with a probability higher than a preset threshold can be recommended to the target customer, or the top n products in terms of probability can be recommended to the target customer. S130: If the target customer is a customer with an interaction history, the target customer's customer attribute information, the product attribute information of each product in the product pool, and the product interaction attribute information are input into the second recommendation model to obtain the probability of purchasing products in the product pool, which is used as the second probability, and product recommendations are made to the target customer based on the second probability.

[0059] For example, product interaction attributes may include information on historical purchases, holdings, and redemptions of the current product, information on the major category to which the historical purchases, holdings, and redemptions of the current product belong, and information on products with the same term as the historical purchases, holdings, and redemptions of the current product.

[0060] In this embodiment of the invention, products with a probability higher than a preset threshold can be recommended to target customers, or products with the highest probability (ranked in the top n) can be recommended to target customers.

[0061] For example, it may also include: summarizing the probabilities of all customers purchasing each product; and for each product, recommending corresponding customers based on the probability of product purchase.

[0062] The method provided in this embodiment may further include summarizing the probability of all customers purchasing each product, and recommending products with high purchase rates to the corresponding customers based on the summarization results. Thus, a product recommendation mechanism is established that simultaneously satisfies both recommending products based on customers and finding customers based on products.

[0063] This embodiment addresses the issue of data skew caused by a limited number of customers purchasing products due to overly detailed product offerings. On one hand, it utilizes customers without interaction history, inputting their customer and product attribute information into the model to determine the probability of purchase. This allows for product recommendations and resolves the cold start problem caused by a lack of customer information. On the other hand, for customers with interaction history, their customer attribute information, product attribute information from the product pool, and product interaction attribute information are input into the model to obtain the probability of purchasing products from the pool. Based on this probability, product recommendations are then made to the customer.

[0064] Example 2

[0065] Figure 2a This is a flowchart illustrating another product recommendation method provided in Embodiment 2 of the present invention. Terms identical to those used in the above embodiments are not repeated here. Figure 2a As shown, the method includes:

[0066] S210. For customers without an interaction history, train the first recommendation model using the first training set; wherein, the first training set includes a first positive sample set and a first negative sample set; the first positive sample set includes feature information corresponding to customers who have no product purchase history and are currently purchasing a product for the first time; the first negative sample set includes feature information corresponding to customers who have no product purchase history and are currently not purchasing a product.

[0067] In this embodiment, since customers' product purchases are influenced by three aspects: customer attribute information (basic attributes and other behaviors within the industry), product attribute information (product category, product cycle, etc.), and product interaction history information, approximately 80% of customers without product interaction history can be modeled separately from customers with product interaction history. The feature information corresponding to customers with no product purchase history who are currently purchasing a product for the first time, as well as the feature information corresponding to customers with no product purchase history who are not currently purchasing a product, are input into the first recommendation model, and the model is trained based on the first recommendation model.

[0068] In this embodiment, a wide table can be constructed based on customers without interaction history and their corresponding customer attribute information and product attribute information. This wide table can be used for model training based on customer attribute information and product attribute information, such as... Figure 2b As shown, this wide table can predict whether a customer will make a purchase based on different customer attribute information and product attribute information. Among them, customer attribute information may include, but is not limited to, current month's payroll, transfer amount, consumption amount, credit card, etc., thereby recommending suitable products to customers based on customer attribute information and product attribute information.

[0069] Therefore, by training the first recommendation model with data corresponding to customers without interaction history, the first recommendation model is obtained. That is, by modeling with data corresponding to customers without interaction history, the impact of customers without interaction history is taken into account, which can solve the cold start problem caused by the lack of customer information.

[0070] S220. For customers with interaction history, train the second recommendation model using the second training set; wherein, the second training set includes a second positive sample set and a second negative sample set; the second positive sample set includes feature information corresponding to customers with product purchase history and currently purchasing products; the second negative sample set includes feature information corresponding to customers with product purchase history and currently not purchasing products; wherein, the feature information in the positive sample set includes customer attribute information, product attribute information, and product interaction attribute information; the feature information in the negative sample set includes customer attribute information and product attribute information.

[0071] In this embodiment, customers with interaction history, along with their corresponding customer attribute information, product attribute information, and product interaction attribute information, can be constructed into a wide table based on the customer's interaction history, such as... Figure 2cAs shown, this wide table can be used to train models based on customer attribute information, product attribute information, and product interaction attribute information. This wide table can predict whether a customer will make a purchase based on different customer attribute information, product attribute information, and product interaction information. Among them, customer attribute information may include, but is not limited to, current month's payroll, transfer amount, consumption amount, and credit card information. Product interaction information may include, but is not limited to, products expiring within one month, number of recent purchases, and purchase amount within the same period. Based on customer attribute information, product attribute information, and product interaction information, suitable products are recommended to customers.

[0072] In this embodiment, a model is created for customers with an interaction history. The feature information of customers with a product purchase history who are currently purchasing a product, as well as the feature information of customers with a product purchase history who are not currently purchasing a product, are input into the second recommendation model, and the model is trained based on the second recommendation model.

[0073] S230. A product pool is formed based on the dimensions of customer attribute similarity, financial attribute similarity, and event marketing.

[0074] In this embodiment, before forming the product pool from the dimensions of customer attribute similarity, financial attribute similarity, and event marketing, it may also include: obtaining customer attribute information, customer behavior data information, product data information, and contract data information; performing feature engineering data processing based on customer attribute information, customer behavior data information, product data information, and contract data information; and obtaining a wide table based on the data processing results and decision tree.

[0075] Customer attribute information may include the customer's payroll company, payroll amount, payroll outlet, customer's financial assets, age, gender, etc.; customer behavior data may include wealth management purchases and current account transactions, etc.; contract data may include mobile banking, online banking, and third-party custody, etc.; feature engineering data processing may include, but is not limited to, data cleaning, normalization, and numerical and discretization processing.

[0076] This embodiment can add typical information about customers such as payroll, mid-to-high-end customers, and credit card customers, such as payroll salary level, industry of payroll customers, customer financial assets and credit cards, thereby specifically forming customer attribute similarity, paying attention to the personalized characteristics of typical customer groups, and thus improving universality.

[0077] In this embodiment, four different wide tables can be obtained. Specifically, a customer attribute similarity wide table can be formed by filtering customer group attributes (payroll, mid-to-high-end, credit card, etc.), basic attributes, and account management attributes to the importance of different product purchases through decision trees; a financial attribute similarity wide table can be formed by filtering customer transaction data, product purchase data, and product holding data to the importance of different product purchases through decision trees; for different bank products, they are classified according to terms and product attributes, and then the similarity between products is calculated by using the Jaccard similarity coefficient based on whether payroll customers jointly hold the products, resulting in a product attribute similarity wide table; and a correlation rule wide table is formed by constructing association rules between customer monthly payroll amount, time, contract signing, transaction, and rights collection information and their current account transaction flow and product purchase and redemption behaviors.

[0078] S240. If the target customer is a customer with no interaction history, input the target customer's customer attribute information and the product attribute information of each product in the product pool into the first recommendation model to obtain the probability of purchasing a product in the product pool, and use it as the first probability, and make product recommendations to the target customer based on the first probability.

[0079] S250. If the target customer is a customer with an interaction history, input the target customer's customer attribute information, the product attribute information of each product in the product pool, and the product interaction attribute information into the second recommendation model to obtain the probability of purchasing a product in the product pool, and use it as the second probability, and make product recommendations to the target customer based on the second probability.

[0080] This embodiment addresses the issue of data skew caused by overly granular product offerings leading to fewer customers making purchases. On the other hand, it models customers with and without interaction history separately to resolve the cold start problem and improve product versatility.

[0081] Example 3

[0082] Figure 3 This is a schematic diagram of a product recommendation device provided in Embodiment 3 of the present invention. Figure 3 As shown, the device includes:

[0083] Product pool formation module 301 is used to form a product pool based on customer attribute similarity, financial attribute similarity, and event marketing.

[0084] The first recommendation module 302 is used to input the customer attribute information of the target customer and the product attribute information of each product in the product pool into the first recommendation model if the target customer is a customer without interaction history, to obtain the probability of purchasing the products in the product pool, and use it as the first probability, and to recommend products to the target customer based on the first probability.

[0085] The second recommendation module 303 is used to input the customer attribute information of the target customer, the product attribute information of each product in the product pool, and the product interaction attribute information into the second recommendation model if the target customer is a customer with an interaction history, to obtain the probability of purchasing a product in the product pool, and use it as a second probability, and to recommend products to the target customer based on the second probability.

[0086] Optionally, the above-mentioned device further includes:

[0087] The training module is used to train the first recommendation model for customers without interaction history using a first training set. The first training set includes a first positive sample set and a first negative sample set. The first positive sample set includes feature information corresponding to customers with no product purchase history who are currently making their first product purchase. The first negative sample set includes feature information corresponding to customers with no product purchase history who are not currently purchasing any products. For customers with interaction history, the second recommendation model is trained using a second training set. The second training set includes a second positive sample set and a second negative sample set. The second positive sample set includes feature information corresponding to customers with product purchase history who are currently purchasing products. The second negative sample set includes feature information corresponding to customers with product purchase history who are not currently purchasing any products. The feature information in the positive sample set includes customer attribute information, product attribute information, and product interaction attribute information. The feature information in the negative sample set includes customer attribute information and product attribute information.

[0088] Optionally, the product pool forming module 301 can be used for:

[0089] In terms of customer attribute similarity, the decision tree filters out the first variable that has a significant impact on customers' product purchases and has no collinearity, based on customer group information, basic attribute information, and account management attribute information.

[0090] The first variable is binned at equal frequency, and the first variable after equal frequency binning is encoded to obtain the first encoded information;

[0091] The first encoded information is input into the Locality Sensitive Hash Model to obtain the customer's first feature vector;

[0092] Products purchased by customers whose feature vectors are similar to the first feature vector within a preset time period are placed into the product pool.

[0093] In terms of financial attribute similarity, based on customers' product purchase information, product holding information, and financial asset information, a second variable that has a significant impact on customers' product purchases and is free from collinearity is selected through decision tree screening.

[0094] The second variable is binned at equal frequency, and the binned second variable is encoded to obtain the second encoded information;

[0095] The second encoded information is input into the Locality Sensitive Hash Model to obtain the customer's second feature vector;

[0096] Products purchased by customers whose feature vectors are similar to the second feature vector within a preset time period are added to the product pool.

[0097] In terms of event marketing, if a customer experiences a target event, the products corresponding to that target event will be added to the product pool.

[0098] Optionally, the customer attribute information includes the first variable;

[0099] The product attribute information includes product sub-category information, product major category information, product term information, product liquidity information, and product profitability information;

[0100] The product interaction attribute information includes information on historical purchases, holdings, and redemptions of the current product; information on the major category to which the current product belongs; and information on products with the same term as the current product.

[0101] Optionally, the target events include product purchase-related events constructed from current account change information, transaction flow information, and product purchase and redemption information.

[0102] Optionally, the product pool forming module 301 can be used for:

[0103] If a customer corresponding to a feature vector similar to the second feature vector purchases the first type of product within a preset time period, and the similarity between the first type of product and the second type of product is higher than a preset threshold, then the first type of product and the second type of product are placed into the product pool.

[0104] If a customer corresponding to a feature vector similar to the second feature vector purchases the first type of product within a preset time period, and the quantity of the first type of product is less than the first preset quantity, then all products in the category to which the first type of product belongs are placed into the product pool.

[0105] If the number of target products purchased by customers corresponding to feature vectors similar to the second feature vector within a preset time period is greater than the second preset number, then under the target product category, the products ranked in the top n by the number of purchasers are placed into the product pool.

[0106] Optionally, the above-mentioned device further includes:

[0107] The summary module is used to summarize the probability of all customers purchasing each product.

[0108] For each product, we recommend corresponding customers based on the probability of purchasing the product.

[0109] The product recommendation device provided in this embodiment of the invention can execute a product recommendation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0110] Example 4

[0111] Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0112] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0113] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0114] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as a product recommendation method.

[0115] In some embodiments, a product recommendation method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of a product recommendation method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform a product recommendation method by any other suitable means (e.g., by means of firmware).

[0116] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0117] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0118] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0119] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0120] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0121] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0122] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0123] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A product recommendation method, characterized in that, include: A product pool is formed based on the dimensions of customer attribute similarity, financial attribute similarity, and event marketing; wherein, the customer attribute similarity is calculated by statistically analyzing the correlation between relevant customer information and products; If the target customer is a customer with no interaction history, the customer attribute information of the target customer and the product attribute information of each product in the product pool are input into the first recommendation model to obtain the probability of purchasing a product in the product pool, which is used as the first probability, and product recommendations are made to the target customer based on the first probability. If the target customer is a customer with an interaction history, the customer attribute information of the target customer, the product attribute information of each product in the product pool, and the product interaction attribute information are input into the second recommendation model to obtain the probability of purchasing products in the product pool, which is used as the second probability, and product recommendations are made to the target customer based on the second probability. The product pool, formed based on customer attribute similarity, financial attribute similarity, and event marketing, includes: In terms of customer attribute similarity, the decision tree filters out the first variable that has a significant impact on customers' product purchases and has no collinearity, based on customer group information, basic attribute information, and account management attribute information. The first variable is binned at equal frequency, and the first variable after equal frequency binning is encoded to obtain the first encoded information; The first encoded information is input into the Locality Sensitive Hash Model to obtain the customer's first feature vector; Products purchased by customers whose feature vectors are similar to the first feature vector within a preset time period are placed into the product pool. In terms of financial attribute similarity, based on customers' product purchase information, product holding information, and financial asset information, a second variable that has a significant impact on customers' product purchases and is free from collinearity is selected through decision tree screening. The second variable is binned at equal frequency, and the binned second variable is encoded to obtain the second encoded information; The second encoded information is input into the Locality Sensitive Hash Model to obtain the customer's second feature vector; Products purchased by customers whose feature vectors are similar to the second feature vector within a preset time period are added to the product pool. In terms of event marketing, if a customer experiences a target event, the product corresponding to the target event will be placed in the product pool; wherein, the target event includes events related to product purchase constructed through current account change information, transaction flow information, and product purchase and redemption information; the target event also includes business experience and business needs; The step of adding products purchased by customers with feature vectors similar to the second feature vector within a preset time period to the product pool includes: If a customer corresponding to a feature vector similar to the second feature vector purchases the first type of product within a preset time period, and the similarity between the first type of product and the second type of product is higher than a preset threshold, then the first type of product and the second type of product are placed into the product pool. If a customer corresponding to a feature vector similar to the second feature vector purchases the first type of product within a preset time period, and the quantity of the first type of product is less than the first preset quantity, then all products in the category to which the first type of product belongs are placed into the product pool. If the number of target products purchased by customers corresponding to feature vectors similar to the second feature vector within a preset time period is greater than the second preset number, then under the target product category, the products ranked in the top n by the number of purchasers are placed into the product pool.

2. The method according to claim 1, characterized in that, Also includes: For customers with no interaction history, the first recommendation model is trained using a first training set; wherein, the first training set includes a first positive sample set and a first negative sample set; the first positive sample set includes feature information corresponding to customers with no product purchase history and who are currently making their first product purchase; the first negative sample set includes feature information corresponding to customers with no product purchase history and who are currently not making a product purchase. For customers with an interaction history, the second recommendation model is trained using the second training set; The second training set includes a second positive sample set and a second negative sample set. The second positive sample set includes feature information corresponding to customers who have product purchase history and are currently purchasing products. The second negative sample set includes feature information corresponding to customers who have product purchase history and are not currently purchasing products. The feature information in the positive sample set includes customer attribute information, product attribute information, and product interaction attribute information. The feature information in the negative sample set includes customer attribute information and product attribute information.

3. The method according to claim 1, characterized in that, Also includes: Sum the probability of all customers purchasing each product; For each product, we recommend corresponding customers based on the probability of purchasing the product.

4. The method according to claim 1, characterized in that, The customer attribute information includes the first variable; The product attribute information includes product sub-category information, product major category information, product term information, product liquidity information, and product profitability information; The product interaction attribute information includes information on historical purchases, holdings, and redemptions of the current product; information on the major category to which the current product belongs; and information on products with the same term as the current product.

5. The method according to claim 1, characterized in that, The target events include product purchase-related events constructed from current account change information, transaction flow information, and product purchase and redemption information.

6. A product recommendation device, characterized in that, include: The product pool formation module is used to form a product pool based on customer attribute similarity, financial attribute similarity, and event marketing; wherein, the customer attribute similarity is calculated by statistically analyzing the correlation between relevant customer information and products; The first recommendation module is used to input the customer attribute information of the target customer and the product attribute information of each product in the product pool into the first recommendation model if the target customer is a customer with no interaction history, to obtain the probability of purchasing a product in the product pool, and use it as the first probability, and to recommend products to the target customer based on the first probability. The second recommendation module is used to input the customer attribute information of the target customer, the product attribute information of each product in the product pool, and the product interaction attribute information into the second recommendation model if the target customer is a customer with an interaction history, to obtain the probability of purchasing a product in the product pool, and use it as the second probability, and to recommend products to the target customer based on the second probability. The product pool formation module is specifically used for: In terms of customer attribute similarity, the decision tree filters out the first variable that has a significant impact on customers' product purchases and has no collinearity, based on customer group information, basic attribute information, and account management attribute information. The first variable is binned at equal frequency, and the first variable after equal frequency binning is encoded to obtain the first encoded information; The first encoded information is input into the Locality Sensitive Hash Model to obtain the customer's first feature vector; Products purchased by customers whose feature vectors are similar to the first feature vector within a preset time period are placed into the product pool. In terms of financial attribute similarity, based on customers' product purchase information, product holding information, and financial asset information, a second variable that has a significant impact on customers' product purchases and is free from collinearity is selected through decision tree screening. The second variable is binned at equal frequency, and the binned second variable is encoded to obtain the second encoded information; The second encoded information is input into the Locality Sensitive Hash Model to obtain the customer's second feature vector; Products purchased by customers whose feature vectors are similar to the second feature vector within a preset time period are added to the product pool. In terms of event marketing, if a customer experiences a target event, the product corresponding to the target event will be placed in the product pool; wherein, the target event includes events related to product purchase constructed through current account change information, transaction flow information, and product purchase and redemption information; the target event also includes business experience and business needs; The product pool forming module is specifically used for: If a customer corresponding to a feature vector similar to the second feature vector purchases the first type of product within a preset time period, and the similarity between the first type of product and the second type of product is higher than a preset threshold, then the first type of product and the second type of product are placed into the product pool. If a customer corresponding to a feature vector similar to the second feature vector purchases the first type of product within a preset time period, and the quantity of the first type of product is less than the first preset quantity, then all products in the category to which the first type of product belongs are placed into the product pool. If the number of target products purchased by customers corresponding to feature vectors similar to the second feature vector within a preset time period is greater than the second preset number, then under the target product category, the products ranked in the top n by the number of purchasers are placed into the product pool.

7. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the method of any one of claims 1-5.

Images