TWM597467U - System for recommending financial product - Google Patents

Abstract

An system for recommending financial product is provided, which updates historical transaction information based on transaction information, and based on the historical transaction information, generates transaction-related information for a plurality of users relating to a plurality of financial products, operation information for a plurality of transaction operations performed by the plurality of users on the financial products, and frequency information of times of which the each user has performed the each transaction operation, and generates user feature matrix and financial product feature matrix based on the operation information and the frequency information, and uses the user feature matrix, the financial product matrix, and the transaction-related information as parameters to establish a recommendation model of the users according to the parameters based on machine learning methods, and uses the recommendation model to provide the users with advertising information related to the plurality of recommendation financial products.

Description

Financial Commodity Recommendation System

This new creation is about a financial commodity recommendation system.

With the rise of convenient financial services, more and more users will compare the credit cards, loan interest rates, and financial commodity interest rates offered by banks, and choose products that are beneficial to them. Therefore, banks often provide users with advertisements for various financial products after logging in to mobile banks. However, advertisements of financial products provided by banks are not necessarily of interest to users who log in to mobile banks. Based on the above, how to provide advertisements of financial commodities that users are more interested in is a subject dedicated to those with ordinary knowledge in the field.

The new creation provides a financial product recommendation system that can provide users with advertisements of financial products that are of interest to users.

The embodiment of the novel creation proposes a financial commodity recommendation system. The financial product recommendation system includes a communication unit, a storage unit, and a processing unit. The communication unit receives transaction information from multiple users. The storage unit stores historical transaction information of multiple users. And the processing unit is connected to the communication unit and the storage unit, and performs the following operations: update historical transaction information based on the transaction information, and generate transaction-related information related to multiple financial commodities by multiple users based on the historical transaction information. The operation information of multiple transaction operations and the number of times that each user performs each transaction operation, where multiple transaction operations are related to multiple financial commodities, and a user feature matrix and a financial product feature matrix are generated based on the operation information and the frequency information , And use the user feature matrix, financial product matrix and transaction related information as parameters, based on the machine learning method to create a recommendation model for multiple users based on the parameters, and use the recommendation model to generate multiple recommendations for multiple users recommended financial products Information to provide advertising information related to recommendation information to multiple users.

In order to make the above-mentioned features and advantages of the creation of the new model more obvious and understandable, the embodiments are specifically described below and described in detail in conjunction with the accompanying drawings.

FIG. 1 is a schematic diagram of a financial commodity recommendation system 100 according to an embodiment of the novel creation. Referring to FIG. 1, the financial product recommendation system 100 has a communication unit 110, a storage unit 120 and a processing unit 130.

The communication unit 110 can receive transaction information of multiple users. In particular, the transaction information may be related to multiple users performing multiple transaction operations, where multiple transaction operations are related to multiple financial commodities. In an embodiment, the transaction information may include the time and consumption amount of each user's transaction operation on any financial commodity, and the account balance after the transaction operation on any financial commodity. For example, the transaction information may include a user among multiple users who spent 10,000 yuan on a credit card on December 30, 2019 and an account balance of 600,000 yuan and a user among multiple users Conduct interbank transfer transactions on December 30, 2019.

In an embodiment of the present invention, the communication unit 110 can be implemented with various types of communication chips. For example, the communication chip can be a global system for mobile communication (GSM), personal handheld phone System (Personal Handy-phone System, PHS), Code Division Multiple Access (CDMA) system, Wideband Code Division Multiple Access (WCDMA) system, Long Term Evolution (LTE) The components of the signal transmission of the system, the Worldwide Interoperability for Microwave Access (WiMAX) system, the Wireless Fidelity (Wi-Fi) system or Bluetooth, but the creation of this new type is not limited to this.

The storage unit 120 is used to store historical transaction information. In an embodiment of the invention, the storage unit 120 is, for example, the storage unit 120 can adopt any type of fixed or removable random access memory (RAM), read-only memory (Read-Only Memory, ROM), flash memory (flash memory), hard disk drive (HDD), solid state drive (SSD) or similar components or a combination of the above components, the new creation is not Limited to this.

The processing unit 130 is connected to the communication unit 110 and the storage unit 120, and updates the historical transaction information stored in the storage unit 120 according to the transaction information. The processing unit 130 is, for example, a central processing unit (Central Processing Unit, CPU), or other programmable general-purpose or special-purpose microprocessor (Microprocessor), a digital signal processor (DSP), or Programmable controllers, application specific integrated circuits (Application Specific Integrated Circuit, ASIC) or other similar components or combinations of the above components, but the creation of the new type is not limited to this.

FIG. 2 illustrates a flowchart of the operation of the financial product recommendation system 100 according to an embodiment of the new creation. Please refer to FIGS. 1 and 2 at the same time. In step S210, the processing unit 130 receives the transaction information through the communication unit 110, and updates the historical transaction information stored in the storage unit 120 according to the transaction information.

In detail, when multiple users log in to the Internet Bank to perform any transaction operation, the processing unit 130 receives transaction information related to multiple transaction operations performed by multiple users on multiple financial products through the communication unit 110, The transaction information is stored in the storage unit 120. In other words, the processing unit 130 records the transaction behavior of each user in the storage unit 120.

In one embodiment, the transaction operation may be credit card consumption, transfer, payment of credit card fees, purchase of foreign exchange, purchase of fixed deposit, etc. In one embodiment, the financial commodities may be deposits, credits, credit cards, foreign exchange, fixed deposits, wealth management, etc. In an embodiment, the processing unit 130 may provide a user interface for each user to use a financial product for trading operations.

In step S220, the processing unit 130 generates transaction-related information related to multiple financial commodities by multiple users, operation information of multiple transaction operations performed by multiple users according to historical transaction information, and each user’s Times information, where multiple trading operations are related to multiple financial commodities.

In one embodiment, the transaction-related information may include the account balance of each user, the financial products purchased by each user in multiple financial products, and the amount spent by each user on each financial product, etc. Limited to this. In one embodiment, the operation information may include making a deposit, using a credit card transaction, performing a transfer transaction, using a credit card to pay a fee, purchasing a fixed deposit, and purchasing foreign exchange, etc., but the creation of the new model is not limited to this. In an embodiment, the processing unit 130 may count the number of times each user performs each transaction operation according to the historical transaction information to generate frequency information.

For example, if the historical transaction information stored by the storage unit 120 includes information that the first user spent 10,000 yuan on a credit card on December 30, 2019, and the account balance was 600,000 yuan, and the second user in 2020 1 On October 10, for inter-bank transfer transactions, the processing unit 130 will determine the transaction-related information based on the above historical transaction information, including the account balance of the first user is 600,000 yuan, the first user has a credit card, and the first user spends one 10,000 yuan, the second user will use the transfer transaction, and determine the operation information based on the above historical transaction information including credit card consumption and transfer transactions, and the number of times of judgment based on the above historical transaction information include the first user using the credit card for one time and the first The second user uses the transfer transaction for one time.

In step S230, the processing unit 130 generates a user feature matrix and a financial product feature matrix based on the operation information and the frequency information, and uses the user feature matrix, financial product matrix, and transaction-related information as parameters, and creates a plurality of based on the parameters based on the machine learning method. The user's recommendation model.

In detail, the processing unit 130 generates a user feature matrix and a financial product feature matrix based on the method of latent factor model in the Collaborative Filtering Recommendation. Further, the processing unit 130 first generates a scoring matrix based on the operation information and the frequency information, where the scoring matrix is a matrix of the number of times each user performs each transaction operation. Next, the processing unit 130 will decompose the scoring matrix into a user feature matrix and a financial commodity feature matrix by using a stochastic gradient descent algorithm to minimize the loss function. In other words, the processing unit 130 will first decompose the scoring matrix into two low-order matrices, and use the stochastic gradient descent algorithm to minimize the loss function of the two low-order matrices to generate a user feature matrix and a financial commodity feature matrix, where The user feature matrix is a matrix of weight values corresponding to each financial product for each user, and the financial product feature matrix is a matrix of weight values corresponding to each transaction operation for each financial product.

For example, if the processing unit 130 determines that x users have performed y transaction operations, the processing unit 130 generates a scoring matrix of x columns and y rows, and determines that the element in the mth column and nth row in the scoring matrix is the mth The number of times the user has performed the nth transaction operation, where x and y are positive integers, m is a positive integer not greater than x, and n is a positive integer not greater than y. In this way, the processing unit 130 can decompose the scoring matrix into low-order matrices, and use a stochastic gradient descent algorithm to minimize the loss function of the two low-order matrices to generate a user feature matrix of x columns and z rows and z columns of y A row of financial product characteristic matrix, where z is a positive integer not greater than y. The elements in the mth column and sth row of the user feature matrix are the weight values of the mth user corresponding to the sth type of financial commodities, where s is a positive integer not greater than z. In other words, as the number of times the mth user purchases the sth type of financial commodities increases, the weight value of the mth user corresponding to the sth type of financial commodities increases. The elements of the nth row of the s column of the financial product feature matrix are the weight values of the sth financial product corresponding to the nth transaction operation. In other words, when the correlation between the sth financial commodity and the nth transaction operation is higher, the weight value of the sth financial commodity corresponding to the nth transaction operation is larger.

Next, the processing unit 130 can use the user feature matrix, financial commodity matrix and transaction-related information as training parameters, and build a multi-class recommendation model based on various machine learning algorithms such as the extreme gradient boosting (eXtreme Gradient Boosting) algorithm (the classification target is Multiple financial products), of which the creation of this new type does not have special restrictions on the machine learning algorithms used. In an embodiment, the processing unit 130 further receives user preference information of multiple users via the communication unit 110, and uses the user feature matrix, transaction classification feature matrix, user preference information, and transaction-related information as training parameters to Based on the above-mentioned parameters, a multi-user recommendation model is established based on the machine learning method, where the user preference information may include the currency preferred by each user, the domestic or foreign transactions preferred by each user, and the transaction operations frequently performed by each user Date etc.

In step S240, the processing unit 130 may use the recommendation model to generate recommendation information of multiple recommended financial products of multiple users to provide multiple users with advertisement information related to the recommended information.

In detail, the processing unit 130 can use the recommendation model to select financial products that may be of interest to multiple users from multiple financial products, and generate advertising information based on the financial products that may be of interest to multiple users. Multiple users provide advertising information related to recommendation information. In one embodiment, in response to multiple users logging into the financial product recommendation system 100 with their respective accounts, the processing unit 130 may provide a user interface to display the above advertisement information.

Based on the above, the financial product recommendation system 100 can provide advertisement information recommending financial products to users who log in to the mobile bank at any time.

In summary, the financial product recommendation system provided by the new creation will obtain various parameters from the stored historical information, and use various parameters as training parameters to generate a recommendation model for multiple users of the mobile bank based on the machine learning method, and then Using this recommendation model, when any one of the multiple users logs in to the mobile bank, advertisement information for recommending financial products is generated for the user who logs in the mobile bank.

Although the new creation has been disclosed as above with examples, it is not intended to limit the creation of the new creation. Any person with ordinary knowledge in the technical field of the subject can make some changes and changes within the spirit and scope of the creation of the new creation. Retouch, so the scope of protection of this new creation shall be subject to the scope defined in the appended patent application.

100: Financial commodity recommendation system 110: communication unit 120: storage unit 130: processing unit S210~S240: Operation steps of financial commodity recommendation system

FIG. 1 is a schematic diagram of a financial commodity recommendation system according to an embodiment of the novel creation. FIG. 2 shows a flowchart of the operation of a financial commodity recommendation system according to an embodiment of the novel creation.

100: Financial commodity recommendation system

110: communication unit

120: storage unit

130: processing unit

Claims (10)

A financial commodity recommendation system, including: Communication unit to receive transaction information from multiple users; A storage unit that stores historical transaction information of the multiple users; and The processing unit connects the communication unit and the storage unit and performs the following operations: Update the historical transaction information according to the transaction information, Generating transaction-related information related to a plurality of financial commodities of the plurality of users, operation information of a plurality of transaction operations performed by the plurality of users, and each of the users performing Information on the number of times of transaction operations, wherein the plurality of transaction operations are related to the plurality of financial commodities, Generate a user feature matrix and a financial product feature matrix based on the operation information and the frequency information, and take the user feature matrix, the financial product matrix, and the transaction-related information as parameters, based on the machine learning method The parameters to establish a recommendation model for the multiple users, and The recommendation model is used to generate recommendation information of a plurality of recommended financial products of the plurality of users, so as to provide advertisement information related to the recommendation information to the plurality of users. The financial product recommendation system according to claim 1, wherein the transaction-related information includes the account balance of each user, the financial products purchased by each user among the plurality of financial products, and each usage The amount of money spent by each of the financial products. The financial product recommendation system according to claim 1, wherein the operation information includes making a deposit, using a credit card transaction, performing a transfer transaction, paying by credit card, purchasing a fixed deposit, and purchasing foreign exchange. The financial commodity recommendation system according to claim 1, wherein the plurality of financial commodities include deposit, credit, credit card, foreign exchange, fixed deposit, and wealth management. The financial commodity recommendation system according to claim 1, wherein the machine learning method is an eXtreme Gradient Boosting algorithm. The financial product recommendation system according to claim 1, wherein the processing unit is generated based on the operation information and the number of times information based on a method of latent factor model in a collaborative filtering algorithm (Collaborative Filtering Recommendation) The user characteristic matrix and the financial commodity characteristic matrix. The financial commodity recommendation system according to claim 6, wherein the step of the processing unit generating the user feature matrix and the financial commodity feature matrix based on the method of the implicit semantic model in the collaborative filtering algorithm includes : Generate a scoring matrix based on the operation information and the number of times information, and use a stochastic gradient descent (Stochastic gradient descent) algorithm to minimize the loss function to decompose the scoring matrix into the user feature matrix and the financial Product feature matrix. The financial product recommendation system according to claim 7, wherein the scoring matrix is a matrix of the number of times each user performs each transaction operation, and the user characteristic matrix is that each user corresponds to each A matrix of weight values of the financial commodities, and the feature matrix of financial commodities is a matrix of weight values of each of the financial commodities corresponding to each of the transaction operations. The financial product recommendation system according to claim 1, wherein the processing unit further receives user preference information of the plurality of users via the communication unit, and uses the user feature matrix and the financial product feature The matrix, the user preference information, and the transaction-related information are used as the parameters, and the recommendation model of the user is established based on the parameters based on the machine learning method, where the user preference information includes each The currency preferred by the user, the domestic or foreign transaction preferred by each user, and the date on which the user often performs transactions. The financial product recommendation system according to claim 1, wherein, in response to the plurality of users logging into the financial product recommendation system with respective accounts, the processing unit provides a user interface to display the advertisement information.

Images