CN115908006A

CN115908006A - Financial product recommendation method, system, equipment and medium based on decision tree

Info

Publication number: CN115908006A
Application number: CN202211349116.7A
Authority: CN
Inventors: 林常乐; 高田金子; 李惟
Original assignee: Cross Information Core Technology Research Institute Xi'an Co ltd
Current assignee: Cross Information Core Technology Research Institute Xi'an Co ltd
Priority date: 2022-10-31
Filing date: 2022-10-31
Publication date: 2023-04-04

Abstract

The invention discloses a financial product recommendation method, a system, equipment and a medium based on a decision tree, which preprocess user data needing to be predicted and financial product data needing to be recommended; establishing time sequence characteristics based on the preprocessed user data needing to be predicted and the preprocessed financial product data needing to be recommended to form a prediction data set; and inputting the prediction data set into a pre-established recommendation model, and outputting a recommended financing product. The invention reserves the time sequence relation of data in the process of data characteristics, and in addition, the decision tree model is adopted as the basic machine learning model, thereby having higher interpretability and low calculation cost.

Description

Financial product recommendation method, system, equipment and medium based on decision tree

Technical Field

The invention relates to the field of financial management, in particular to a method, a system, equipment and a medium for recommending financial products based on a decision tree.

Background

With the development of big data, the financial industry has more profound understanding on the application value of information technology, and the machine learning model is applied to various aspects of the financial industry, such as helping financial institutions make loan decisions, anti-money laundering supervision, accurate marketing and the like. The precise marketing is based on precise positioning, and a personalized customer communication service system is established by means of modern information technology, and personalized recommendation is an important part of the precise marketing.

The personalized recommendation of the financial product refers to recommending a proper product to a proper user at a proper time. On one hand, from the perspective of financial institution users, it is more meaningful to receive the information of the financial products which are interested by the users than to receive the products which are not interested by the users or cannot be purchased; on the other hand, from the perspective of financial institutions, the financial product information and suggestions which are most likely to be purchased by the user are recommended to the user, so that the user stickiness can be improved, and the income can be increased. The financial institution completes the task of personalized recommendation by establishing and debugging a recommendation model based on data such as user information, account state, historical purchase records and the like.

Common recommendation models and algorithms include collaborative filtering, matrix decomposition, and deep learning network-based recommendation algorithms. The former two are more classical recommendation algorithms, the latter being a very popular approach for this few years. However, the financial products are different from general physical commodities in nature, and the types of the financial products which can be selected are less than those of the physical commodities. The decision of the user to purchase the financial product is often greatly related to the previous purchasing behavior, for example, if the previous financial product is good in income, the user will purchase the same or similar product with a high probability next time. Therefore, when building a model, we need to consider the time-series relationship in the historical purchase data of the user, that is, the model needs to be able to process the time-series data or to gain insight into the relationship.

The more traditional recommendation algorithm is represented by collaborative filtering and matrix decomposition recommendation. Collaborative filtering is a recommendation algorithm designed based on user behavior, and specifically finds some similarity (e.g., similarity between items or similarity between items) through group behaviors, and essentially makes decisions and recommendations for users through the similarity. The matrix decomposition recommendation algorithm is to arrange the user data into a user-product matrix, wherein each item in the matrix represents the preference of a certain user for a certain product and has a certain missing value. The data matrix is then decomposed into the product of the two most likely low rank matrices by a matrix decomposition technique. And then, the product of the two low-rank matrixes is used for reversely deducing the missing in the user-product matrix, so that the product recommendation is carried out.

The other recommendation system framework is characterized in that a deep neural network model is used as a core, and neural networks with different structures are constructed for recommendation. For data with timing relationships, the most commonly considered used is the Long Short-term Memory model (LSTM). The LSTM model is a variant of a recurrent neural network, and the long-term dependence problem of the recurrent neural network is solved by adding gate control (namely when the time sequence is too long, the recurrent neural network model forgets information of a longer time period, and the more recent time point has greater influence on the input at the moment). The recommendation algorithm based on the simple LSTM model takes the historical behavior data of the user as input, the purchased products are labels, and a product which is predicted to be most likely purchased by the user at the next time point can be obtained after training.

The traditional classical recommendation method, collaborative filtering and matrix decomposition, on one hand, cannot effectively process the time dependence relationship between data. However, a large part of data of financial product users is historical behavior data, and the data cannot be avoided to have time sequence. And through data mining research, the purchasing behavior of the user at a certain time point is strongly correlated with the historical behavior of the user. On the other hand, both methods, although highly explanatory, may be difficult to land in practice, especially in the case of a large number of users. The collaborative filtering requires a large amount of computing resources, and the matrix decomposition recommendation algorithm may face the cold start problem, that is, the historical behavior of a certain user is too little to be predicted.

On the other hand, the recommendation method based on the LSTM model can effectively process time series data and obtain good model effect. However, as with other deep neural networks, on the one hand, a large amount of computational resources are required to train the model. On the other hand, the network structure is complex, the interpretability is poor, and the model correctness is difficult to verify. In the financial industry, however, interpretability is a very important requirement, and in addition to the requirement that the model perform well, the model also needs to be reasonably constructed to meet the common financial knowledge.

Disclosure of Invention

The invention aims to provide a financial product recommendation method, a system, equipment and a medium based on a decision tree, aiming at overcoming the defects in the prior art, the recommendation method adopted by the invention aims at the financial product, takes the recommendation specificity of the financial product into consideration, and on one hand, the processing of data characteristics keeps the time sequence relation of data; on the other hand, a decision tree model is finally adopted as a basic machine learning model, and the model has high interpretability and low calculation cost.

In order to achieve the purpose, the invention adopts the following technical scheme:

the financial product recommendation method based on the decision tree comprises the following steps:

preprocessing user data needing to be predicted and financial product data needing to be recommended;

establishing time sequence characteristics based on the preprocessed user data needing to be predicted and the preprocessed financial product data needing to be recommended to form a prediction data set;

inputting the prediction data set into a pre-established recommendation model, and outputting a recommended financing product;

the establishment method of the recommendation model specifically comprises the following steps:

acquiring historical user data and historical financial product data;

preprocessing historical user data and historical financial product data;

establishing a time sequence characteristic based on the preprocessed historical user data and the historical financial product data to form a historical data set;

and training and testing the decision tree model by using the historical data set to obtain a recommendation model.

Further, the preprocessing of the historical financial product data specifically comprises:

and extracting the characteristics of the historical financing product data, and outputting a financing product information dictionary, wherein the financing product information dictionary takes the characteristics of the financing product as a label, and stores the serial number and the corresponding name of the financing product.

Further, the preprocessing the historical user data specifically includes:

and carrying out missing value processing and abnormal value processing on historical user data, wherein the historical user data comprises historical transaction records of historical users and historical user account information.

Further, the establishing of the time sequence characteristics based on the preprocessed historical user data and the historical financial product data specifically comprises:

and establishing a time sequence characteristic containing a time relation based on the preprocessed historical financial product data and historical user data and by combining historical user account information, historical user transaction records and historical financial product information.

Further, the specific method for establishing the time sequence characteristics and forming the historical data set comprises the following steps:

assuming that the financial products most likely to be purchased by the user at the prediction time node are to be predicted;

based on the predicted time node, for each user, forward intercepting data based on a preset time window, and for each time window, extracting statistical characteristics corresponding to the intercepted data;

and combining the statistical characteristics with the user information to predict financial products purchased by the time node to be used as a label to form a historical data set.

Further, the statistical features include mean and median.

Further, the training and testing of the decision tree model by using the historical data set to obtain the recommendation model specifically comprises:

dividing a historical data set into a training connection set and a test set;

training a decision tree model by utilizing a training set;

and evaluating the result of the trained decision tree model by using the test set, and if the result does not reach the expected result, adjusting the decision tree model until the evaluation reaches the expected result to obtain a recommended model.

The financial product recommendation system based on decision tree includes:

a preprocessing module: the system is used for preprocessing user data needing to be predicted and financial product data needing to be recommended;

a predictive dataset formation module: the system comprises a data processing module, a data processing module and a data processing module, wherein the data processing module is used for establishing time sequence characteristics based on preprocessed user data needing to be predicted and preprocessed financial product data needing to be recommended to form a prediction data set;

a recommendation module: the system is used for inputting the prediction data set into a pre-established recommendation model and outputting a recommended financing product;

acquiring historical user data and historical financial product data;

preprocessing historical user data and historical financial product data;

A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the decision tree based financial product recommendation method when executing the computer program.

A computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the decision tree based financial product recommendation method.

Compared with the prior art, the invention has the following beneficial technical effects:

the recommendation method adopted by the invention aims at financial products, considers the specificity of recommendation of financial products, and on one hand, the processing of data characteristics keeps the time sequence relation of data; on the other hand, a decision tree model is finally adopted as a basic machine learning model, and the model has high interpretability and low calculation cost.

Furthermore, the recommendation method adopted by the invention aims at financial products, and intercepts different time windows and extracts data information during data feature processing so as to keep the time sequence relation of the data. The processed data contains time sequence relation, but the data is not time sequence data, so that a simpler machine learning model can be adopted for modeling. In order to meet the requirement of the financial field on model interpretability, the machine learning model with higher interpretability, namely the decision tree, is used as a basic model, and the model can meet the requirements of interpretability and accuracy.

Drawings

The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.

FIG. 1 is a flow chart of the present invention for building a recommendation model;

FIG. 2 is a flow chart of the present invention using a recommendation model.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Example one

The invention provides a recommendation method for financial products, which is used for realizing the recommendation of financial products and considering the characteristics of financial products, so that the time sequence relation can be processed on one hand, and the adopted machine learning model is simple enough and easy to understand on the other hand.

The invention can be divided into two large modules: a recommendation model is established (before market entry) and applied (after market entry). The process of establishing the recommendation model comprises the following steps: 1) Inputting data (historical user data and historical financial product data) and preprocessing the data; 2) Selecting a model according to requirements and specific scenes, and training the model; 3) And performing model evaluation, if the model is poor in performance, readjusting the model, and if the model is good in performance, outputting the model as a recommended model to be applied.

Putting the trained model into the market for use, wherein the part of the general process is as follows: 1) Inputting user data to be predicted and financing product data to be recommended to perform data preprocessing; 2) Predicting by using the established recommendation model; 3) And (4) outputting the recommended financing product, feeding back the model effect and the recommendation result to the modeling process, and preparing for model adjustment iteration.

Specifically, the recommendation method provided by the invention has the following flows:

modeling section, as shown in fig. 1:

step 0: confirming partial parameters in advance, and establishing time windows of the time sequence characteristics, such as 3 months and 6 months; considering a recommendation system, the top K most likely financial products to be purchased are recommended.

Step 1: and (4) preprocessing the data of the historical financing product, extracting the characteristics of the historical financing product, and outputting a product information dictionary (roughly comprising product labels, characteristics and the like).

Financial products for financial institutions are generally complex in name, for example, over time, there may be a range of products named XXXX1, XXXX2, \8230, 8230, which are substantially identical in character but differ in name. If all financial products in the user purchase record are marked, on one hand, the situation of excessive labels is caused; on the other hand, a situation that a new label does not appear before occurs, and the model cannot identify the new label in prediction.

Therefore, the invention performs characteristic extraction on the financial products, such as closed/non-closed products, closed product time, product series names and the like. And the characteristics of the financing product are taken as a label, and the product number and the corresponding product name are stored.

And 2, step: and preprocessing historical user data.

The historical user transaction history and the historical user account information (if any) are preprocessed in a conventional preprocessing manner, including but not limited to missing value processing and abnormal value processing.

And step 3: and establishing time sequence characteristics, and generating a historical data set for the recommendation model.

Based on the preprocessed historical financial product data (step 1) and historical user data (step 2), the time sequence characteristics including the time relationship are established by combining the historical user account information, the historical user transaction records, the historical financial product information and other data, and the specific method for establishing the time sequence characteristics comprises the following steps:

with months as a time measure, assume that the financial products that the user is most likely to purchase at the prediction time node t are to be predicted (selected in the historical data, which are actually known).

Based on the predicted time node, for each user, data is intercepted forward based on a preset time window. For example, the time windows are 3 months and 6 months, and then 3 months (t-3 to t-1) and 6 months (t-6 to t-1) of historical data are taken forward with the prediction time point as the node. Then, for each time window, extracting the statistical characteristics corresponding to each characteristic, such as the mean value, the median and the like.

And (4) combining the statistical characteristics with the user information, and generating a historical data set by taking the financial products purchased at the predicted time point as labels.

And 4, step 4: training and testing models, outputting recommendation models

Using the historical data set generated in the step 3, dividing the historical data set into a training set and a testing set;

the model is established through training of a training set, wherein a decision tree model is adopted in the method, and the method has strong interpretability and visibility;

evaluating the model result by using the test set, and adjusting the model if the model result is not expected;

and outputting the recommendation model.

Model application part, as shown in fig. 2:

step 0: and step 0 of the modeling portion;

step 1: preprocessing the data of the financial product to be predicted, extracting the characteristics of the financial product, combining the previous product information dictionary, and adding the product into the dictionary according to the characteristics if the product cannot be in the dictionary.

And 2, step: the pre-processing of the user data, which requires prediction, is similar to step 2 of the modeling section.

And step 3: and establishing a time sequence characteristic, and generating a data set for prediction.

The method is similar to step 3 of the modeling section, except that this section has no label because the purchasing behavior at the new point in time is predicted.

And 4, step 4: and inputting the prediction data set into a recommendation model, and outputting a recommendation result of the recommendation model.

Example two

The invention provides a financial product recommendation system based on decision trees, which comprises the following components:

acquiring historical user data and historical financial product data;

preprocessing historical user data and historical financial product data;

EXAMPLE III

The invention also provides a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the decision tree-based financial product recommendation method when executing the computer program.

Example four

The present invention also provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the steps of the financial product recommendation method.

Interpretation of terms

A decision tree model: and a tree structure is adopted, and layer-by-layer reasoning is used for realizing final classification or prediction.

Time series data: refer to time series data.

Training set, testing set: when building a machine learning model, datasets are generally divided into two categories: the training set is used for training the machine learning model (building an optimal model based on the training set), and the test set is used for measuring the performance of the trained model.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

Finally, it should be noted that: although the present invention has been described in detail with reference to the above embodiments, those skilled in the art will appreciate that various changes, modifications and equivalents can be made in the embodiments of the invention without departing from the scope of the invention as defined by the appended claims.

Claims

1. The financial product recommendation method based on the decision tree is characterized by comprising the following steps:

acquiring historical user data and historical financial product data;

preprocessing historical user data and historical financial product data;

2. The decision tree-based financial product recommendation method according to claim 1, wherein the preprocessing of the historical financial product data is specifically:

3. The decision tree-based financial product recommendation method according to claim 2, wherein the preprocessing of the historical user data is specifically:

and carrying out missing value processing and abnormal value processing on historical user data, wherein the historical user data comprises historical user transaction records and historical user account information.

4. The decision tree-based financial product recommendation method according to claim 3, wherein the establishing of the time sequence feature based on the preprocessed historical user data and the historical financial product data specifically comprises:

5. The decision tree-based financial product recommendation method according to claim 4, wherein the specific method for establishing the time series characteristics and forming the historical data set is as follows:

and combining the statistical characteristics with the user information to predict financial products purchased by the time node as a label to form a historical data set.

6. The decision tree-based financial product recommendation method according to claim 5, wherein the statistical features include mean and median.

7. The decision tree-based financial product recommendation method according to claim 5, wherein the decision tree model is trained and tested using a historical data set to obtain a recommendation model, specifically:

dividing a historical data set into a training set and a testing set;

training a decision tree model by utilizing a training set;

8. Decision tree-based financial product recommendation system, comprising:

a predictive dataset formation module: the system comprises a data processing unit, a data processing unit and a data processing unit, wherein the data processing unit is used for establishing time sequence characteristics based on preprocessed user data needing to be predicted and preprocessed financial product data needing to be recommended to form a prediction data set;

acquiring historical user data and historical financial product data;

preprocessing historical user data and historical financial product data;

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the decision tree based financial product recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the steps of the decision tree based financial product recommendation method according to any one of claims 1 to 7.