Financial product recommendation method, computer device and computer readable storage medium

Abstract

The invention provides a financial product recommendation method, a computer device and a computer readable storage medium, and the method comprises the steps: building an enterprise database, obtaining a target static data vector of a target enterprise, calculating the distance between the target static data vector and static data vectors of remaining enterprises in the enterprise database, obtaining a first remaining enterprise, and obtaining a second remaining enterprise; forming a similar enterprise set; financial products historically applied by the first remaining enterprise are obtained, financial products in the second sorting list are obtained, and a first recommended financial set is formed; inputting the financial product historical application vector of the target enterprise and the financial product historical application vectors of the remaining enterprises in the enterprise database into a collaborative filtering model, and outputting a second recommended financial product set; and forming a combined financial product set, sorting the combined financial product set according to the response time of each financial product, and outputting a recommended financial product set. According to the invention, financial products with higher recommendation accuracy are recommended.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent recommendation of financial products, and in particular to a method for recommending financial products, a computer device and a computer-readable storage medium. Background Art

[0002] When existing enterprises need to choose financial products, they often ask banks to recommend relevant financial products. Banks usually recommend financial products based on questionnaires or basic information filled out by enterprises, or use machine learning and big data analysis technologies to analyze and recommend. Financial products recommended through bank questionnaires are recommended by bank staff after analysis. Since it is difficult to consider various information about enterprises during manual analysis, the financial products recommended manually are not accurate and may not meet the needs of enterprises. Although technologies such as machine learning and big data analysis can use more data analysis, they often only analyze based on some basic information of enterprises, making it difficult to achieve the most accurate recommendation. The data recommended by technologies such as machine learning and big data analysis may not meet the needs of enterprises.

[0003] An existing digital intelligent matching method for financial products for small and micro enterprises is to establish a label library of enterprise user loan needs, enterprise user information, and financial product information to form a label portrait of enterprise users and financial products; based on the labels of various dimensions of enterprise users and financial product information labels, a recall strategy is used to accurately screen financial products; a logistic regression algorithm is used to train and evaluate the pass rate of each product; the financial products initially screened by the recall strategy are sorted according to the pass rate and interest rate, and recommended to enterprise users. However, this matching method only analyzes the static data of the enterprise and makes recommendations without considering the behavioral characteristics of the enterprise. The recommended financial products may not meet the needs of the enterprise, and the experience of enterprise users is poor. Summary of the invention

[0004] The first object of the present invention is to provide a method for recommending financial products with higher recommendation accuracy.

[0005] A second object of the present invention is to provide a computer device for implementing the above-mentioned financial product recommendation method.

[0006] A third object of the present invention is to provide a computer-readable storage medium for applying the above-mentioned financial product recommendation method.

[0007] To achieve the first objective of the present invention, the present invention provides a financial product recommendation method, which includes: establishing an enterprise database, the enterprise database includes multiple enterprises, and one enterprise includes a static data vector and a financial product history application vector; obtaining a target static data vector of a target enterprise, calculating the distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database, sorting the remaining enterprises from near to far according to the distance to obtain a first sorted list, obtaining the first first preset number of first remaining enterprises in the first sorted list, and forming a similar enterprise set; obtaining financial products historically applied for by the first remaining enterprise from the similar enterprise set, sorting the multiple financial products from most to least according to the frequency of occurrence to obtain a second sorted list, and forming a first recommended financial set by obtaining the first second preset number of financial products in the second sorted list; obtaining the financial product history application vector of the target enterprise, inputting the financial product history application vector of the target enterprise and the financial product history application vector of the remaining enterprises in the enterprise database into a collaborative filtering model, and outputting a second recommended financial product set; merging the first recommended financial product set with the second recommended financial product set to form a merged financial product set, sorting the merged financial product set according to the response time of each financial product, and outputting the recommended financial product set.

[0008] It can be seen from the above scheme that the financial products in the recommended financial product set include historical application financial products of similar enterprises obtained by comparing static data vectors, and also include historical application financial products of similar enterprises obtained by comparing historical application vectors. Static data vectors can reflect the similarity of the overall attributes between the target enterprise and similar enterprises, so the historical application financial products of similar enterprises meet the needs of the target enterprise. Since there is a collaborative filtering calculation by comparing the historical application vectors of the financial products of the target enterprise and the historical application vectors of the financial products of the remaining enterprises, it is possible to accurately find enterprises with similar historical application behaviors as the target enterprise, so that more suitable financial products can be recommended to the target enterprise. Among them, the historical application of financial products of enterprises is a dynamic feature. The dynamic features of multiple aspects such as credit situation, corporate reputation and capital operation ability of enterprises with similar historical application behaviors are relatively similar, but similar historical application behaviors not only reflect the dynamic features of credit situation, corporate reputation and capital operation ability. Make the products recommended by the financial product set more in line with the requirements of the enterprise. Find similar enterprises through static data vectors and historical application financial product vectors, find similar enterprises from the static and dynamic features of enterprises, so that the recommended financial product set is more in line with the needs of the target enterprise. Financial products are sorted by their response time and those with short response time are recommended to target enterprises first, thereby enhancing the user experience of target enterprises.

[0009] In a further scheme, the step of establishing an enterprise database includes: calculating the static data vector of each enterprise, the static data vector includes a first static data vector and a second static data vector; the calculation parameters of the first static data vector and the second static data vector include at least one of the following: enterprise registered capital, high-tech enterprise, specialized and new enterprise, employee social security payment level, average operating income; the first static data vector also includes expected loan amount, expected interest rate, and expected term; calculating the financial product historical application vector of each enterprise, and the calculation parameters of the enterprise financial product historical application vector include financial product historical application data.

[0010] It can be seen that the static data vector takes into account factors such as the size, registered capital, and industrial attributes of the enterprise, and personalizes the recommended financial products to meet the different needs of the enterprise.

[0011] In a further scheme, the step of obtaining the target static data vector of the target enterprise includes: determining whether the expected loan amount, expected interest rate, and expected term of the first static data vector of the target enterprise are all zero; if so, obtaining the second static data vector of the target enterprise as the static data vector of the target enterprise; if the expected loan amount, expected interest rate, and expected term of the first static data vector of the target enterprise are not all zero, then obtaining the first static data vector of the target enterprise as the target static data vector of the target enterprise.

[0012] It can be seen that through the loan term, expected interest rate and expected term, the system can achieve accurate matching of financial products and provide the most suitable financial products for the enterprise. If the enterprise does not provide data such as loan term, expected interest rate and expected term, the data cannot be used for calculation, so the second static data vector is required for calculation.

[0013] In a further solution, the step of calculating the distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database includes: calculating the cosine distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database.

[0014] In a further scheme, after obtaining the first preset number of first remaining enterprises in the first sorted list, the following is also performed: the first preset number of first remaining enterprises in the first sorted list are formed into a first similar enterprise set; the comprehensive distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database is calculated, and sorted from near to far according to the comprehensive distance to obtain a third sorted list, and the first third preset number of second remaining enterprises in the third sorted list are obtained, and it is determined whether multiple second remaining enterprises are the same as the first remaining enterprise; if the second remaining enterprise is the same as the first remaining enterprise, the first remaining enterprise in the first similar enterprise set is retained; if not, the first remaining enterprise that is different from the second remaining enterprise is removed from the first similar enterprise set; and a similar enterprise set is obtained.

[0015] In a further solution, the comprehensive distance is calculated by cosine distance and Euclidean distance; the Euclidean distance is calculated by the distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database.

[0016] It can be seen that the comprehensive evaluation through cosine distance and Euclidean distance comprehensively considers the directional similarity and spatial similarity of enterprise basic data, thereby improving the accuracy of recommendation.

[0017] In a further scheme, after the merged financial product set is formed, the following is also performed: judging whether the enterprise is a specialized, sophisticated, innovative enterprise or a high-tech enterprise; if so, arranging the financial products related to the specialized, sophisticated, innovative enterprise or the high-tech enterprise in the merged financial product set to the starting position of the merged financial product set.

[0018] It can be seen from this that financial products related to specialized, sophisticated, innovative enterprises or high-tech enterprises should be placed at the forefront of consolidated financial products to meet the needs of specialized, sophisticated, innovative enterprises or high-tech enterprises.

[0019] In order to achieve the second purpose, the computer device provided by the present invention includes a processor and a memory, the memory stores a computer program, and when the computer program is executed by the processor, the method for recommending financial products mentioned above is implemented.

[0020] In order to achieve the third purpose, the present invention provides a computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is executed, the above-mentioned financial product recommendation method is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of an embodiment of a method for recommending financial products.

[0022] The present invention is further described below in conjunction with the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0023] The financial product recommendation method provided by the present invention obtains financial products by selecting similar enterprises through target static data vectors and financial product historical application vectors, and separately searches for similar enterprises through the static data and dynamic data of the enterprises, so that the obtained recommended financial product set better meets the needs of users.

[0024] Implementation example of financial product recommendation method: See also Figure 1 The financial product recommendation method of this embodiment is implemented by a computer device. First, step S1 is executed to establish an enterprise database. The enterprise database includes multiple enterprises, and each enterprise includes a static data vector and a financial product history application vector. The step of establishing the enterprise database includes: calculating the static data vector of each enterprise; calculating the financial product history application vector of each enterprise. Preferably, the enterprise database needs to be updated once a day.

[0025] Among them, the static data vector includes a first static data vector and a second static data vector, and the calculation parameters of the first static data vector and the second static data vector include at least one of the following: enterprise registered capital, high-tech enterprises, specialized and new enterprises, employee social security payment level, and average operating income.

[0026] When calculating the static data vectors of each enterprise, the calculation parameters of the first static data vector and the second static data vector must first be preprocessed. The enterprise registered capital preprocessing is the enterprise registered capital parameter. The enterprise registered capital parameter is ; Among them, regiterCapital is the registered capital. High-tech enterprise preprocessing is high-tech enterprise parameters, high-tech enterprise parameters are , where 0 means the enterprise is not a high-tech enterprise, 1 means the enterprise was once a high-tech enterprise, and 2 means the enterprise is a high-tech enterprise. , where 0 represents that the enterprise is not a specialized, refined, special and new enterprise, and 2 represents that the enterprise is a specialized, refined, special and new enterprise. The employee social security payment level preprocessing is the employee social security payment level parameter, and the employee social security payment level parameter is avgPeopleCount is the average social security payment level of each employee in the past 12 months. The average operating income is preprocessed into the average operating income parameter. The average operating income parameter is , avgIncome is the average monthly operating income in the past 12 months.

[0027] The first static data vector also includes the expected loan amount, the expected interest rate, and the expected term. When preprocessing the data, the expected loan amount is preprocessed into the expected loan amount parameter. , expectedAmount is the expected loan amount. The expected interest rate is preprocessed into the expected interest rate parameter, which is , expectedInterest is the expected interest. The expected term is preprocessed into the expected term parameter, and the expected term parameter is . Among them, expectedTerm is the expected term.

[0028] Since the static data vector includes a first static data vector and a second static data vector, the first static data vector is , the second static data vector is .

[0029] The calculation parameter of the financial product historical application vector is the company's financial product historical application data. When calculating the financial product historical application vector of each company, the financial product historical application data is recorded using one-hot encoding. The steps for calculating the financial product historical application vector of each company include: obtaining a list of financial products, generating a primary key for the financial products, and sorting the financial products from large to small according to the primary key. When calculating a financial product historical application vector, the company's financial product historical application data is obtained, where the financial product record that has passed the bank's initial review is 1, and the rest of the records are 0. The company's financial product historical application vector is obtained as Among them, the financial product historical application vector is sorted in descending order according to the primary key of the financial product.

[0030] After the enterprise database is established, step S2 is executed to obtain the target static data vector of the target enterprise and calculate the distance between the target static data vector and the static data vectors of the remaining enterprises, wherein the remaining enterprises are all enterprises in the enterprise database except the target enterprise.

[0031] Among them, the step of obtaining the target static data vector of the target enterprise includes: determining whether the expected loan amount, expected interest rate, and expected term of the first static data vector of the target enterprise are all zero; if so, obtaining the second static data vector of the target enterprise as the static data vector of the target enterprise; if the expected loan amount, expected interest rate, and expected term of the first static data vector of the target enterprise are not all zero, then obtaining the first static data vector of the target enterprise as the target static data vector.

[0032] Calculate the distance between the target static data vector and the static data vectors of the remaining enterprises. The distance is the cosine distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database. The calculation formula of the cosine distance is: Among them, V (1) is the target static data vector, V (2) is a static data vector for one of the remaining enterprises.

[0033] After calculating the distance between the target static data vector and the static data vector of the remaining enterprises, execute step S13 to sort the remaining enterprises according to the cosine distance to obtain a first sorted list, and obtain the first preset number of first remaining enterprises in the first sorted list, that is, obtain the first preset number of first remaining enterprises with the closest distance to form a set of similar enterprises.

[0034] Among them, after obtaining the first preset number of first remaining enterprises in the first sorting list, it is also necessary to form the first similar enterprise set with the first preset number of first remaining enterprises in the first sorting list; calculate the comprehensive distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database. Among them, the calculation formula of the comprehensive distance is: the comprehensive distance is calculated by the cosine distance and the Euclidean distance, and the Euclidean distance is calculated by the distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database. The calculation formula of the Euclidean distance is: , V (1) is the target static data vector, V (2) is the static data vector of one of the remaining enterprises. The calculation formula of the comprehensive distance is . Where i is the index of the remaining enterprises in the enterprise database.

[0035] After calculating the comprehensive distance between the target static data vector and the static data vector of the remaining enterprises in the enterprise database, sort them from near to far according to the comprehensive distance to obtain a third sorting list. Obtain the first third preset number of second remaining enterprises in the third sorting list, and determine whether the plurality of second remaining enterprises are the same as the first remaining enterprise; if the second remaining enterprise is the same as the first remaining enterprise, retain the first remaining enterprise in the first similar enterprise set; if the second remaining enterprise is different from the first remaining enterprise, remove the first remaining enterprise that is different from the second remaining enterprise from the first similar enterprise set to obtain a similar enterprise set.

[0036] After obtaining the similar enterprise set, step S14 is executed to obtain the financial products historically applied for by the first remaining enterprise from the similar enterprise set, sort the multiple financial products according to the frequency of occurrence, obtain a second sorted list, obtain the first second preset number of financial products in the second sorted list, and form a first recommended set. The financial products in the first recommended set exclude the financial products historically applied for by the target enterprise.

[0037] After forming the first recommendation set, step S15 is executed to obtain the target enterprise's historical application vector for financial products, input the target enterprise's historical application vector for financial products and the remaining enterprises in the enterprise database into the collaborative filtering model, and output the second recommended financial product set. The collaborative filtering model uses a collaborative filtering algorithm, which recommends the remaining enterprises with similar application behaviors to the target enterprise, and then outputs a set of financial products that the remaining enterprises have applied for but the target enterprise has not applied for. The first fourth preset number of financial products is obtained from the financial product set to form the second recommended financial product set.

[0038] After forming the second recommended financial product set, execute step S16 to merge the first recommended financial product set with the second recommended financial product set to form a merged financial product set. After forming the merged financial product set, determine whether the enterprise is a specialized, sophisticated, innovative enterprise or a high-tech enterprise; if so, arrange the financial products of the specialized, sophisticated, innovative enterprise or high-tech enterprise in the merged financial product set to the starting position of the merged financial product set. Sort the merged financial product set according to the response time of each financial product, and output the recommended financial product set. The response time is the bank's response time to the financial product from pending to pending review or rejection. Sorting by response time improves the enterprise user experience.

[0039] This embodiment performs static data analysis on the target enterprise and historical data analysis on financial products to find similar enterprises with higher accuracy from static and dynamic aspects, thereby recommending financial products that better meet the requirements of the target enterprise.

[0040] Computer device embodiment: The computer device of this embodiment includes a processor and a memory. The memory stores a computer program. When the processor executes the computer program, the above-mentioned financial product recommendation method is implemented.

[0041] The computer device may include but is not limited to a processor and a memory. Those skilled in the art will appreciate that the computer device may include more or fewer components, or a combination of certain components, or different components, for example, the computer device may also include input and output devices, network access devices, buses, etc.

[0042] Computer readable storage medium embodiment: The financial product recommendation method in the computer device described in the above embodiment can be stored in a computer-readable storage medium in the form of a computer program. When the computer program is executed by a processor, the steps of the financial product recommendation method embodiment in the computer device can be completed. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0043] The above are only preferred embodiments of the present invention, but the design concept of the invention is not limited thereto. Without departing from the concept of the present invention, more other equivalent embodiments may be included. It is obvious to those skilled in the art that various changes, readjustments and substitutions can be made without departing from the protection scope of the present invention.

Claims

1. A method for recommending financial products, characterized in that: The method includes: Establishing an enterprise database, wherein the enterprise database includes a plurality of enterprises, and each enterprise includes a static data vector and a financial product history application vector; Obtain a target static data vector of a target enterprise, calculate the distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database, sort the remaining enterprises from near to far according to the distance to obtain a first sorted list, obtain a first preset number of first remaining enterprises in the first sorted list, and form a similar enterprise set; Obtain financial products historically applied for by the first remaining enterprise from the similar enterprise set, sort the multiple financial products from most to least according to the frequency of occurrence to obtain a second sorted list, obtain the first second preset number of financial products in the second sorted list, and form a first recommended financial set; Obtaining a historical application vector of a financial product of a target enterprise, inputting the historical application vector of the financial product of the target enterprise and the historical application vectors of the financial product of the remaining enterprises in the enterprise database into a collaborative filtering model, and outputting a second recommended financial product set; The first recommended financial product set and the second recommended financial product set are combined to form a combined financial product set, the combined financial product set is sorted according to the response time of each financial product, and the recommended financial product set is output.

2. The method for recommending financial products according to claim 1, characterized in that: The steps of establishing the enterprise database include: Calculating a static data vector for each enterprise, wherein the static data vector includes a first static data vector and a second static data vector; The calculation parameters of the first static data vector and the second static data vector include at least one of the following: enterprise registered capital, high-tech enterprise, specialized and new enterprise, employee social security payment level, and average operating income; The first static data vector also includes an expected loan amount, an expected interest rate, and an expected term; The financial product historical application vector of each enterprise is calculated, and the calculation parameters of the financial product historical application vector include the financial product historical application data.

3. The method for recommending financial products according to claim 2, characterized in that: The steps of obtaining the target static data vector of the target enterprise include: Determining whether the expected loan amount, expected interest rate, and expected term of the first static data vector of the target enterprise are all zero; If yes, obtaining a second static data vector of the target enterprise as the static data vector of the target enterprise; If the expected loan amount, expected interest rate, and expected term of the first static data vector of the target enterprise are not all zero, then the first static data vector of the target enterprise is obtained as the target static data vector.

4. The method for recommending financial products according to any one of claims 1 to 3, characterized in that: The step of calculating the distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database comprises: The cosine distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database is calculated.

5. The method for recommending financial products according to claim 4, characterized in that: After obtaining the first preset number of first remaining enterprises in the first sorting list, further executing: The first preset number of first remaining enterprises in the first sorting list form a first similar enterprise set; Calculate the comprehensive distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database, sort the remaining enterprises from near to far according to the comprehensive distance to obtain a third sorted list, obtain the first third preset number of second remaining enterprises in the third sorted list, and determine whether the plurality of second remaining enterprises are the same as the first remaining enterprise; If the second remaining enterprise is the same as the first remaining enterprise, the first remaining enterprise in the first set of similar enterprises is retained; If not, the first remaining enterprise that is different from the second remaining enterprise is removed from the first similar enterprise set; The similar enterprise set is obtained.

6. The method for recommending financial products according to claim 5, characterized in that: The comprehensive distance is calculated by the cosine distance and the Euclidean distance; The Euclidean distance is calculated from the distance between the target static data vector and the static data vectors of the remaining enterprises in the enterprise database.

7. The method for recommending financial products according to any one of claims 1 to 3, characterized in that: After forming a combined financial product set, the following is also performed: Determine whether the enterprise is a specialized, sophisticated, innovative or high-tech enterprise; If so, arrange the financial products related to the specialized, innovative enterprises or high-tech enterprises in the merged financial product set to the starting position of the merged financial product set.

8. A computer device comprising a processor and a memory, characterized in that The memory stores a computer program, and when the computer program is executed by the processor, the method for recommending financial products according to any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for recommending financial products according to any one of claims 1 to 7 is implemented.

Images