Financial product recommendation method and device based on user personal information

Abstract

The invention relates to the technical field of financial product recommendation, and discloses a financial product recommendation method and device based on user personal information, and the method comprises the steps: inputting the user information into a financial product recommendation system, importing related data into a CNN model, carrying out the feature extraction of the imported data, and outputting a first feature tag; the method comprises the following steps: inputting historical transaction data (arranged according to a time sequence) of a user into an LSTM model, processing time sequence data by using an LSTM unit, outputting first data, splicing first features output by a CNN model and the LSTM model with the first data to form a comprehensive feature representation, namely, a second feature tag, and storing the second feature tag in the CNN model; the spliced second features are mapped to probability distribution of recommended products through a full connection layer, and the probability that each financial product is recommended is output by using a softmax function. The related information of the user is input into the financial product recommendation system embedded in the computer body, the financial product recommendation scheme is quickly generated, and the scheme is imported into the tablet personal computer of the user side.

Description

Technical Field

[0001] This invention relates to the field of financial product recommendation technology, specifically to a method and apparatus for recommending financial products based on user personal information. Background Technology

[0002] With the development of the Internet and the arrival of the big data era, people have gradually moved from an era of information scarcity to an era of information overload. Recommendation systems have emerged as a result. On the one hand, recommendation systems can help users efficiently obtain the financial products they need from a vast array of financial products, improving the user experience of online platforms. On the other hand, they increase the click-through rate and user stickiness of online platforms.

[0003] For example, a financial product recommendation method and apparatus with Chinese patent number CN118505350A obtains the latest financial products browsed by the user and searches for the latest financial products in the user's historical browsing financial product database. The historical browsing financial product database stores the relationship between the various historical financial products browsed by the user in the past. The related historical financial products are similar. If the historical browsing financial product database includes the latest financial product, the method obtains the historical financial products in the historical browsing financial product database that are associated with the latest financial product and pushes the historical financial products associated with the latest financial product to the user terminal.

[0004] However, the aforementioned recommendation mechanism can only recommend financial products that users have browsed, which is limited and cannot help customers avoid corresponding risks. As a result, the recommendation scheme has poor practicality and cannot recommend personalized schemes based on the user. The selection is low, resulting in poor recommendation effect. Therefore, a financial product recommendation method and device based on user personal information is proposed to solve the above-mentioned problems. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a method and apparatus for recommending financial products based on user personal information. This method offers advantages such as greater practicality and more personalized solutions. It solves the problems of existing recommendation mechanisms, which can only recommend financial products that users have already viewed, resulting in limited options and an inability to mitigate risks for customers. Consequently, the recommendation solutions are less practical and cannot be personalized based on individual user preferences, leading to low selectivity and poor recommendation performance.

[0007] Technical solution

[0008] To achieve the aforementioned goals of enhanced practicality and more personalized solutions, this invention provides the following technical solution: a method for recommending financial products based on user personal information, characterized by comprising the following steps:

[0009] Step 1: Collect users' personal information, historical transaction data, financial product information, market data and other relevant information, preprocess the data and construct effective features to be extracted;

[0010] Step 2: Build a CNN convolutional neural network to extract the corresponding features from Step 1, including user financial product features, market data features, etc.

[0011] Step 3: Use a deep learning framework to build an LSTM model and use LSTM units to process time series data, namely, the purchase and browsing data of financial products by users within a preset time period.

[0012] Step 4: Embed both the CNN model and the LSTM model from Steps 2 and 3 into the product recommendation system;

[0013] Step 5: Import the data from the database into the CNN model, perform feature extraction on the imported data through convolutional layers, and output the first feature label;

[0014] Step Six: Input the user's historical transaction data from the database into the LSTM model, process the time series data using LSTM units, and output the first data.

[0015] Step 7: Concatenate the first features and first data from the outputs of the CNN model and the LSTM model to form a comprehensive feature representation, i.e., the second feature label, making the features more specific and accurate;

[0016] Step 8: Map the concatenated second feature onto the probability distribution of the recommended product through the fully connected layer of the CNN model;

[0017] Step 9: Use the softmax function to output the probability of each financial product being recommended, and compare it with the actual needs to adjust the model parameters to ensure that the recommendations better meet the needs;

[0018] Step 10: Recommend 3-4 products with the highest probability ranking to the user for the customer to choose from.

[0019] Preferably, the financial product information includes yield, risk level and category, and the user's personal information includes age, gender and historical transaction data.

[0020] Preferably, the data preprocessing includes cleaning the data, handling missing and outlier values, converting categorical data into numerical data, and performing normalization or standardization, thereby making the data more complete.

[0021] Preferably, the CNN convolutional neural network model uses the following steps:

[0022] S1 first inputs the preprocessed data into the CNN model;

[0023] S2 uses multiple convolutional kernels to extract different local features;

[0024] S3 then reduces the feature dimension and improves computational efficiency by using max pooling or average pooling;

[0025] S4 maps the pooled features to a fixed-dimensional vector, which represents the probability distribution of the recommended products.

[0026] Preferably, the LSTM model data processing steps include:

[0027] 1) Input the user's historical transaction data into the LSTM;

[0028] 2) Use LSTM units to process time series data, including users' past transaction data and browsing data within a preset time period, thereby capturing users' long-term dependency relationships;

[0029] 3) Finally, the data is output through the output layer. The output of LSTM can be a representation of the user's preferences or needs for future financial products.

[0030] A financial product recommendation device based on user personal information includes an apparatus for a financial product recommendation method based on user personal information, comprising a computer body and a tablet computer, wherein the computer body has a financial product recommendation system embedded therein, and the computer body and the tablet computer are wirelessly connected.

[0031] Preferably, the financial product recommendation system includes a financial data storage module, a data processing module, a data splicing module, a probability calculation module, and a result output module.

[0032] Preferably, the data processing module includes a CNN convolutional neural network feature extraction module and an LSTM time-series data extraction module.

[0033] Preferably, the data storage module stores user's historical transaction data, financial product information such as yield, risk level or category, and market data such as index trends;

[0034] The data processing module includes the extraction of data features and the temporal arrangement of data, which facilitates the matching of user information with the financial products.

[0035] The data splicing module splices the data output from the first financial feature output module and the first time series data output module to form the second feature extraction data.

[0036] The probability calculation module maps the second feature to the probability distribution of the recommended products and calculates the probability of each financial product being recommended using the softmax function.

[0037] Beneficial effects

[0038] Compared with the prior art, the present invention provides a method and apparatus for recommending financial products based on user personal information, which has the following beneficial effects:

[0039] 1. This financial product recommendation method based on user personal information combines CNN and LSTM models through steps one through ten, integrating the advantages of convolutional neural networks and long short-term memory networks to more accurately predict user interests and preferences, thereby making the recommendations more personalized and practical.

[0040] 2. This financial product recommendation device based on user personal information inputs the user's relevant information into the financial product recommendation system embedded in the computer, quickly generates a financial product recommendation plan, and imports the plan into the user's tablet computer, so that the user can quickly select a suitable recommendation plan, thereby achieving a more efficient and convenient purpose. Attached Figure Description

[0041] Figure 1 This is a flowchart of the recommended method for the product of the present invention;

[0042] Figure 2 This is a schematic diagram of the recommended device for the product of the present invention.

[0043] In the diagram: 1. Computer body, 2. Financial product recommendation system, 3. Tablet computer. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0045] Please see Figure 1-2 A method for recommending financial products based on user personal information, characterized by the following steps:

[0046] Step 1: Collect users' personal information, historical transaction data, financial product information, market data and other relevant information, preprocess the data and construct effective features to be extracted;

[0047] Step 2: Build a CNN convolutional neural network to extract the corresponding features from Step 1, including user financial product features, market data features, etc.

[0048] Step 3: Use a deep learning framework to build an LSTM model and use LSTM units to process time series data, namely, the purchase and browsing data of financial products by users within a preset time period.

[0049] Step 4: Embed both the CNN and LSTM models from Steps 2 and 3 into the product recommendation system;

[0050] Step 5: Import the data from the database into the CNN model, perform feature extraction on the imported data through convolutional layers, and output the first feature label;

[0051] Step Six: Input the user's historical transaction data from the database into the LSTM model in time series order, process the time series data using LSTM units, and output the first data.

[0052] Step 7: Concatenate the first features and first data from the outputs of the CNN model and the LSTM model to form a comprehensive feature representation, i.e., the second feature label, making the features more specific and accurate;

[0053] Step 8: Map the concatenated second feature onto the probability distribution of the recommended product through the fully connected layer of the CNN model;

[0054] Step 9: Use the softmax function to output the probability of each financial product being recommended, and compare it with the actual needs to adjust the model parameters to ensure that the recommendations better meet the needs;

[0055] Step 10: Recommend 3-4 products with the highest probability ranking to the user for the customer to choose from.

[0056] exist Figure 1 In the system, financial product information includes yield, risk level and category, while user personal information includes age, gender and historical transaction data.

[0057] Specifically, by setting up step one, a database is established. The massive amount of data can support the model, enabling the training of CNN and LSTM models, and allowing for continuous parameter adjustments and updates to ensure more accurate recommendations that better meet customer needs.

[0058] exist Figure 1 In this process, data preprocessing includes cleaning the data, handling missing and outlier values, converting categorical data into numerical data, and performing normalization or standardization, thereby making the data more complete and easier to input.

[0059] exist Figure 1 In this context, the CNN convolutional neural network model uses the following steps:

[0060] S1 first inputs the preprocessed data, such as financial product characteristics and market data, into the CNN model;

[0061] S2 uses multiple convolutional kernels to extract different local features;

[0062] S3 then reduces the feature dimension and improves computational efficiency by using max pooling or average pooling;

[0063] S4 maps the pooled features to a fixed-dimensional vector, which represents the probability distribution of the recommended products.

[0064] Specifically, multiple features are extracted, and pooling is used to reduce computational load, thereby improving the response speed of the CNN model and increasing recommendation efficiency.

[0065] exist Figure 1 In LSTM model data processing, the steps include:

[0066] 1) Input the user's historical transaction data into the LSTM in a time series order;

[0067] 2) Use LSTM units to process time series data, including users' past transaction data and browsing data within a preset time period, thereby capturing users' long-term dependency relationships;

[0068] 3) Finally, the data is output through the output layer. The output of LSTM can be a representation of the user's preferences or needs for future financial products.

[0069] Specifically, the LSTM model can better predict subsequent financial product purchases based on users' transaction behavior within a preset time period, thus making it easier to make accurate product recommendations.

[0070] exist Figure 2 The system includes a computer body 1 and a tablet computer 3. The computer body 1 has a financial product recommendation system 2 embedded inside it, and the computer body 1 and the tablet computer 3 are wirelessly connected.

[0071] The financial product recommendation system 2 includes a financial data storage module, a data processing module, a data splicing module, a probability calculation module, and a result output module.

[0072] The system includes: a data storage module that stores users' historical transaction data, financial product information such as yield, risk level or category, and market data such as index trends; a data processing module that extracts data features and arranges the data in a time sequence to facilitate matching user information with the financial products; a data splicing module that splices the data output from the first financial feature output module and the first time-series data output module to form second feature extraction data; and a probability calculation module that maps the second features to the probability distribution of recommended products and calculates the probability of each financial product being recommended using the softmax function.

[0073] Specifically, through the collaborative work of various modules, the financial regulation system is made more robust, responds faster, and recommends financial products that better meet user needs.

[0074] In summary, this method and apparatus for recommending financial products based on user personal information, through step one, divides the data in the database into training and validation sets, and trains the CNN model through knowledge distillation to ensure more accurate feature extraction by the CNN model. Through steps two and three, the CNN model and LSTM model are combined to accurately predict the current market situation, product trends, transaction rates, etc., and combine them with the customer's past transaction habits, so that the recommended financial products are more personalized and better meet the user's needs.

[0075] Furthermore, the user information of the product to be recommended is first input into the financial product recommendation system 2, and the relevant data is imported into the CNN model. The imported data is then processed by convolutional layers to extract features and output the first feature label. The user's historical transaction data (arranged in time series) is then input into the LSTM model, and the LSTM unit processes the time series data and outputs the first data. The first features and the first data output from the CNN model and the LSTM model are concatenated together to form a comprehensive feature representation, namely the second feature label. The concatenated second feature is then mapped onto the probability distribution of the recommended products through a fully connected layer. The softmax function is used to output the probability of each financial product being recommended, and the top 3-4 products with the highest probability are recommended to the user for selection.

[0076] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for recommending financial products based on user personal information, characterized in that: Includes the following steps: Step 1: Collect users' personal information, historical transaction data, financial product information, market data and other relevant information, preprocess the data and construct effective features to be extracted; Step 2: Build a CNN convolutional neural network to extract the corresponding features from Step 1, including user financial product features, market data features, etc. Step 3: Use a deep learning framework to build an LSTM model and use LSTM units to process time series data, namely, the purchase and browsing data of financial products by users within a preset time period. Step 4: Embed both the CNN model and the LSTM model from Steps 2 and 3 into the product recommendation system; Step 5: Import the data from the database into the CNN model, perform feature extraction on the imported data through convolutional layers, and output the first feature label; Step 6: Input the user's historical transaction data (arranged in time series) from the database into the LSTM model, process the time series data using LSTM units, and output the first data. Step 7: Concatenate the first features and first data from the outputs of the CNN model and the LSTM model to form a comprehensive feature representation, i.e., the second feature label, making the features more specific and accurate; Step 8: Map the concatenated second feature onto the probability distribution of the recommended product through the fully connected layer of the CNN model; Step 9: Use the softmax function to output the probability of each financial product being recommended, and compare it with the actual needs to adjust the model parameters to ensure that the recommendations better meet the needs; Step 10: Recommend 3-4 products with the highest probability ranking to the user for the customer to choose from.

2. The method for recommending financial products based on user personal information according to claim 1, characterized in that: The financial product information includes yield, risk level and category, and the user's personal information includes age, gender and historical transaction data.

3. The method for recommending financial products based on user personal information according to claim 1, characterized in that: The data preprocessing includes cleaning the data, handling missing and outlier values, converting categorical data into numerical data, and performing normalization or standardization, thereby making the data more complete.

4. The method for recommending financial products based on user personal information according to claim 1, characterized in that: The CNN convolutional neural network model uses the following steps: S1 first inputs the preprocessed data (such as financial product characteristics, market data, etc.) into the CNN model; S2 uses multiple convolutional kernels to extract different local features; S3 then reduces the feature dimension and improves computational efficiency by using max pooling or average pooling; S4 maps the pooled features to a fixed-dimensional vector, which represents the probability distribution of the recommended products.

5. The method for recommending financial products based on user personal information according to claim 1, characterized in that: The LSTM model data processing steps include: 1) Input the user's historical transaction data (arranged in time series) into the LSTM; 2) Use LSTM units to process time series data, including users' past transaction data and browsing data within a preset time period, thereby capturing users' long-term dependency relationships; 3) Finally, the data is output through the output layer. The output of LSTM can be a representation of the user's preferences or needs for future financial products.

6. A financial product recommendation device based on user personal information, comprising an apparatus for executing the financial product recommendation method based on user personal information according to any one of claims 1-5, characterized in that: It includes a computer body (1) and a tablet computer (3). The computer body (1) has a financial product recommendation system (2) embedded inside it. The computer body (1) and the tablet computer (3) are wirelessly connected.

7. A financial product recommendation device based on user personal information according to claim 6, characterized in that: The financial product recommendation system (2) includes a financial data storage module, a data processing module, a data splicing module, a probability calculation module, and a result output module.

8. A financial product recommendation device based on user personal information according to claim 6, characterized in that: The data processing module includes a CNN convolutional neural network feature extraction module and an LSTM time series data extraction module.

9. A financial product recommendation device based on user personal information according to claim 6, characterized in that: The data storage module stores user's historical transaction data, financial product information such as yield, risk level or category, and market data such as index trends. The data processing module includes the extraction of data features and the temporal arrangement of data, which facilitates the matching of user information with the financial products. The data splicing module splices the data output from the first financial feature output module and the first time series data output module to form the second feature extraction data. The probability calculation module maps the second feature to the probability distribution of the recommended products and calculates the probability of each financial product being recommended using the softmax function.

Images