Financial account security prediction system and method based on multi-dimensional features
By combining multi-dimensional feature analysis and models, a financial account security prediction system is constructed, which solves the problems of single data dimension and poor adaptability of traditional financial account security prediction, and realizes efficient identification and risk control of new fraud behaviors.
Patent Information
- Application Number
- CN202510783138.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-12
- Publication Date
- 2025-09-12
AI Technical Summary
Traditional financial account security prediction methods rely on a single data dimension, ignore device environment and user behavior, and are unable to adapt to the dynamic changes in financial business scenarios, resulting in a decrease in the accuracy and effectiveness of security predictions, and poor adaptability to different types of accounts and transaction scenarios.
We use multi-dimensional feature analysis, including basic account information, transaction behavior, device environment, user behavior, and external related data, combined with random forest, gradient boosting tree, and LSTM models, to build a security prediction system, provide real-time warnings, and take risk control measures.
It significantly improves the accuracy and comprehensiveness of risk identification, can timely prevent new types of fraud, reduce the risk of financial loss, and improve risk management efficiency.
Smart Images
Figure CN120634699A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of financial account security prediction, and in particular to a financial account security prediction system and method based on multi-dimensional features. Background Art
[0002] As the digitalization of the financial industry accelerates, the security threats facing financial accounts are becoming increasingly complex and diverse. Traditional financial account security prediction methods are gradually failing to meet real-world needs. Existing security predictions rely heavily on account transaction data, judging risk based on only a few dimensions such as transaction amount and time, while ignoring key data such as device environment, user behavior, and external related information. Without considering the security status of devices and abnormal changes in user login locations, potential risks cannot be fully captured, resulting in a large number of new fraud behaviors being difficult to identify in a timely manner.
[0003] Traditional models use fixed risk assessment rules based on historical data, making them incapable of adapting to dynamic changes in financial business scenarios, such as business model innovation and changes in user behavior. Furthermore, they are unable to adjust prediction strategies in a timely manner in the face of new cyberattacks and fraudulent techniques, significantly reducing the accuracy and effectiveness of security predictions.
[0004] At the same time, when dealing with financial account security prediction problems, single models or simple combination models have poor adaptability to different types of accounts (savings, credit cards, investment accounts, etc.) and diversified transaction scenarios. They cannot fully utilize the complex characteristics and patterns in the data, and it is difficult to ensure stable prediction performance in different financial institutions or business scenarios.
[0005] Therefore, a financial account security prediction system and method based on multi-dimensional features are proposed. Summary of the Invention
[0006] Technical problems solved
[0007] In response to the shortcomings of the existing technology, the present invention provides a financial account security prediction system and method based on multi-dimensional features, which has the advantages of multi-dimensional precise analysis and efficient risk warning, and solves the above-mentioned problems of single data dimension, lack of dynamic adaptability and weak model generalization ability.
[0008] Technical Solution
[0009] To achieve the above-mentioned object, the present invention provides the following technical solution: a financial account security prediction system based on multi-dimensional features, comprising a central processing unit, wherein a receiving end of the central processing unit is signal-connected to a transmitting end of a server;
[0010] The central processing unit receiving end is signal-connected to the data preparation unit transmitting end, the central processing unit receiving end is signal-connected to the model training unit transmitting end, and the central processing unit receiving end is signal-connected to the security prediction and response unit transmitting end;
[0011] The receiving end of the security prediction and response unit is signal-connected to the transmitting end of the model training unit, the receiving end of the model training unit is signal-connected to the transmitting end of the data preparation unit, and the receiving end of the data preparation unit is signal-connected to the transmitting end of the security prediction and response unit.
[0012] Preferably, the data preparation unit includes a data acquisition module, a data transmission module and a data processing module, the transmitting end of the data acquisition module is signal-connected to the receiving end of the data transmission module, and the transmitting end of the data transmission module is signal-connected to the receiving end of the data processing module;
[0013] The data collection module includes an account basic information collection module, a transaction behavior data collection module, a device environment data collection module, a user behavior data collection module and an external related data collection module.
[0014] Preferably, the data processing module includes a data cleaning module, a data conversion module, a data feature engineering operation module and a multi-dimensional feature data set construction module;
[0015] The data cleaning module transmitting end is signal-connected to the data conversion module receiving end, the data conversion module transmitting end is signal-connected to the data feature engineering operation module receiving end, and the data feature engineering operation module transmitting end is signal-connected to the multi-dimensional feature data set construction module receiving end.
[0016] Preferably, the model training unit includes a security prediction model module and a feature data set division module, and the feature data set division module includes a training set module, a validation set module and a test set module;
[0017] The training set module transmitting end is signal-connected to the model training module receiving end;
[0018] The validation set module transmitter is signal-connected to the model evaluation module receiver, the model evaluation module transmitter is signal-connected to the model optimization module receiver, and the model optimization module transmitter is signal-connected to the model hyperparameter adjustment module receiver;
[0019] The test set module transmitting end is signal-connected to the model verification module receiving end.
[0020] Preferably, the safety prediction and response unit includes a model real-time prediction module, a model output prediction risk level module and a model visualization display module, and the transmitting end of the model output prediction risk level module is signal-connected to the receiving end of the risk warning signal sending module.
[0021] Another technical problem to be solved by the present invention is to provide a prediction method for a financial account security prediction system based on multi-dimensional features, comprising the following steps:
[0022] Step 1: Use the data collection module to obtain basic account information, transaction behavior data, device environment data, user behavior data, and external related data from various data sources;
[0023] Step 2: The data collected by the data acquisition module is transmitted to the data processing module through the data transmission module;
[0024] Step 3: After receiving the data, the data processing module performs data cleaning, data conversion, and feature engineering operations in sequence to build a complete multi-dimensional feature data set;
[0025] Step 4: Select the random forest, gradient boosting tree, and LSTM models, and determine the initial parameter settings for the security prediction model module;
[0026] Step 5: Use the feature dataset partitioning module to divide the feature dataset obtained in step 3 into a training set, a validation set, and a test set;
[0027] Step 6: Use the training set to train the model. During the training process, adjust the model hyperparameters based on the evaluation results of the validation set to optimize the model performance. Finally, use the test set to evaluate the model with the final parameters to verify the model's generalization ability and prediction accuracy.
[0028] Step 7: During system operation, newly generated account operation and transaction data is received in real time, pre-processed, and then input into the trained model;
[0029] Step 8: When the model outputs the predicted risk level as medium or high risk through the model output prediction risk level module, the risk warning process is triggered and the warning information is sent to relevant personnel through the risk warning signal sending module.
[0030] Preferably, the step eight further includes, after receiving the warning information, the bank security management department taking corresponding security measures for the risky accounts according to the warning information and recording the processing results.
[0031] Preferably, the security measures taken include restricting transactions and requiring identity verification.
[0032] Compared with the existing technology, the present invention provides a financial account security prediction system and method based on multi-dimensional features, which has the following beneficial effects:
[0033] 1. This multi-dimensional feature-based financial account security prediction system and method comprehensively characterizes account security characteristics by collecting multi-dimensional data such as basic account information, transaction behavior, device environment, user behavior, and external connections. By combining feature engineering based on time, space, and behavioral patterns, it can deeply explore potential risks, significantly improve the accuracy and comprehensiveness of risk identification, and effectively prevent new types of fraud.
[0034] 2. This financial account security prediction system and method based on multi-dimensional features can trigger multi-channel warnings for predicted medium- and high-risk operations through the system, quickly notifying account holders and financial institution security departments, facilitating timely risk control measures and reducing the risk of capital loss. The visual display function helps security managers intuitively grasp the risk situation and improve risk management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 This is a schematic diagram of the overall structural connection of the present invention;
[0036] Figure 2 This is a schematic diagram of the data preparation unit structure connection of the present invention;
[0037] Figure 3 This is a schematic diagram of the structure connection of the model training unit of the present invention;
[0038] Figure 4 This is a schematic diagram of the structural connection of the security prediction and response unit of the present invention.
[0039] In the figure: central processing unit 1, server 2, data preparation unit 3, data acquisition module 31, data transmission module 32, data processing module 33, model training unit 4, security prediction model module 41, feature data set division module 42, security prediction and response unit 5. DETAILED DESCRIPTION
[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0041] See also Figure 1-Figure 4 ;
[0042] The financial account security prediction system based on multi-dimensional features includes a central processing unit 1, a receiving end of the central processing unit 1 and a transmitting end of a server 2 connected to each other;
[0043] The receiving end of the central processing unit 1 is signal-connected to the transmitting end of the data preparation unit 3, the receiving end of the central processing unit 1 is signal-connected to the transmitting end of the model training unit 4, and the receiving end of the central processing unit 1 is signal-connected to the transmitting end of the security prediction and response unit 5;
[0044] The receiving end of the security prediction and response unit 5 is signal-connected to the transmitting end of the model training unit 4, the receiving end of the model training unit 4 is signal-connected to the transmitting end of the data preparation unit 3, and the receiving end of the data preparation unit 3 is signal-connected to the transmitting end of the security prediction and response unit 5.
[0045] The data preparation unit 3 includes a data acquisition module 31, a data transmission module 32 and a data processing module 33. The transmitting end of the data acquisition module 31 is signal-connected to the receiving end of the data transmission module 32, and the transmitting end of the data transmission module 32 is signal-connected to the receiving end of the data processing module 33.
[0046] The data collection module 31 includes an account basic information collection module, a transaction behavior data collection module, a device environment data collection module, a user behavior data collection module, and an external related data collection module;
[0047] The data processing module 33 includes a data cleaning module, a data conversion module, a data feature engineering operation module and a multi-dimensional feature data set construction module;
[0048] The transmitting end of the data cleaning module is signal-connected to the receiving end of the data conversion module, the transmitting end of the data conversion module is signal-connected to the receiving end of the data feature engineering operation module, and the transmitting end of the data feature engineering operation module is signal-connected to the receiving end of the multi-dimensional feature data set construction module.
[0049] The model training unit 4 includes a security prediction model module 41 and a feature data set division module 42. The feature data set division module 42 includes a training set module, a verification set module and a test set module.
[0050] The transmitter of the training set module is connected to the receiver of the model training module;
[0051] The transmitter of the validation set module is connected to the receiver of the model evaluation module, the transmitter of the model evaluation module is connected to the receiver of the model optimization module, and the transmitter of the model optimization module is connected to the receiver of the model hyperparameter adjustment module;
[0052] The transmitting end of the test set module is connected to the receiving end of the model verification module;
[0053] The safety prediction and response unit 5 includes a model real-time prediction module, a model output prediction risk level module and a model visualization display module. The transmitting end of the model output prediction risk level module is signal-connected to the receiving end of the risk warning signal sending module.
[0054] Another technical problem to be solved by the present invention is to provide a prediction method for a financial account security prediction system based on multi-dimensional features, comprising the following steps:
[0055] Step 1: Use the data collection module 31 to obtain basic account information, transaction behavior data, device environment data, user behavior data and external related data from various data sources;
[0056] Step 2: The data collected by the data collection module 31 is transmitted to the data processing module 33 through the data transmission module 32;
[0057] Step 3: After receiving the data, the data processing module 33 performs data cleaning, data conversion and feature engineering operations in sequence to build a complete multi-dimensional feature data set;
[0058] Step 4: Select the random forest, gradient boosting tree, and LSTM models and determine the initial parameter settings for the security prediction model module 41;
[0059] Step 5: Using the feature data set division module 42, the feature data set obtained in step 3 is divided into a training set, a validation set, and a test set;
[0060] Step 6: Use the training set to train the model. During the training process, adjust the model hyperparameters based on the evaluation results of the validation set to optimize the model performance. Finally, use the test set to evaluate the model with the final parameters to verify the model's generalization ability and prediction accuracy.
[0061] Step 7: During system operation, newly generated account operation and transaction data is received in real time, pre-processed, and then input into the trained model;
[0062] Step 8: When the model outputs a predicted risk level of medium or high risk through the model output prediction risk level module, the risk warning process is triggered, and a warning message is sent to relevant personnel through the risk warning signal sending module; when the bank's security management department receives the warning message, it takes corresponding security measures for risky accounts based on the warning message, including restricting transactions and requiring identity verification, and records the processing results.
[0063] The beneficial effects of the present invention are as follows: This multi-dimensional feature-based financial account security prediction system and method comprehensively characterizes account security characteristics by collecting multi-dimensional data such as basic account information, transaction behavior, device environment, user behavior, and external connections. By combining feature engineering with temporal, spatial, and behavioral patterns, it can deeply explore potential risks, significantly improve the accuracy and comprehensiveness of risk identification, and effectively prevent new types of fraud. The system instantly triggers multi-channel warnings for predicted medium- and high-risk operations, quickly notifying account holders and financial institution security departments, facilitating the timely implementation of risk control measures and reducing the risk of financial loss. The visualization function helps security managers intuitively grasp the risk landscape and improve risk management efficiency.
[0064] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.
Claims
1. A financial account security prediction system based on multi-dimensional features, comprising a central processing unit (1), characterized in that: The central processing unit (1) receiving end is connected to the server (2) transmitting end by signal; The receiving end of the central processing unit (1) is connected to the transmitting end of the data preparation unit (3) by signal, the receiving end of the central processing unit (1) is connected to the transmitting end of the model training unit (4) by signal, and the receiving end of the central processing unit (1) is connected to the transmitting end of the security prediction and response unit (5) by signal; The receiving end of the security prediction and response unit (5) is signal-connected to the transmitting end of the model training unit (4), the receiving end of the model training unit (4) is signal-connected to the transmitting end of the data preparation unit (3), and the receiving end of the data preparation unit (3) is signal-connected to the transmitting end of the security prediction and response unit (5).
2. The financial account security prediction system based on multi-dimensional features according to claim 1 is characterized in that: The data preparation unit (3) includes a data acquisition module (31), a data transmission module (32) and a data processing module (33), wherein the transmitting end of the data acquisition module (31) is signal-connected to the receiving end of the data transmission module (32), and the transmitting end of the data transmission module (32) is signal-connected to the receiving end of the data processing module (33); The data collection module (31) includes an account basic information collection module, a transaction behavior data collection module, a device environment data collection module, a user behavior data collection module and an external related data collection module.
3. The financial account security prediction system based on multi-dimensional features according to claim 2 is characterized in that: The data processing module (33) includes a data cleaning module, a data conversion module, a data feature engineering operation module and a multi-dimensional feature data set construction module; The data cleaning module transmitting end is signal-connected to the data conversion module receiving end, the data conversion module transmitting end is signal-connected to the data feature engineering operation module receiving end, and the data feature engineering operation module transmitting end is signal-connected to the multi-dimensional feature data set construction module receiving end.
4. The financial account security prediction system based on multi-dimensional features according to claim 1 is characterized in that: The model training unit (4) includes a security prediction model module (41) and a feature data set partitioning module (42), wherein the feature data set partitioning module (42) includes a training set module, a verification set module, and a test set module; The training set module transmitting end is signal-connected to the model training module receiving end; The validation set module transmitter is signal-connected to the model evaluation module receiver, the model evaluation module transmitter is signal-connected to the model optimization module receiver, and the model optimization module transmitter is signal-connected to the model hyperparameter adjustment module receiver; The test set module transmitting end is signal-connected to the model verification module receiving end.
5. The financial account security prediction system based on multi-dimensional features according to claim 1 is characterized in that: The safety prediction and response unit (5) includes a model real-time prediction module, a model output prediction risk level module and a model visualization display module. The model output prediction risk level module transmitting end is signal-connected to the risk warning signal sending module receiving end.
6. The prediction method of the financial account security prediction system based on multi-dimensional features according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Using the data acquisition module (31) to obtain basic account information, transaction behavior data, device environment data, user behavior data and external related data from various data sources; Step 2: transmitting the data collected by the data collection module (31) to the data processing module (33) via the data transmission module (32); Step 3: After receiving the data, the data processing module (33) performs data cleaning, data conversion and feature engineering operations in sequence to construct a complete multi-dimensional feature data set; Step 4: Select random forest, gradient boosting tree and LSTM models and determine the initial parameter settings of the security prediction model module (41); Step 5: Using the feature data set partitioning module (42), the feature data set obtained in step 3 is divided into a training set, a validation set, and a test set; Step 6: Use the training set to train the model. During the training process, adjust the model hyperparameters based on the evaluation results of the validation set to optimize the model performance. Finally, use the test set to evaluate the model with the final parameters to verify the model's generalization ability and prediction accuracy. Step 7: During system operation, newly generated account operation and transaction data is received in real time, pre-processed, and then input into the trained model; Step 8: When the model outputs the predicted risk level as medium or high risk through the model output prediction risk level module, the risk warning process is triggered and the warning information is sent to relevant personnel through the risk warning signal sending module.
7. The method for predicting financial account security based on multi-dimensional features according to claim 6, characterized in that: The step eight also includes, after receiving the early warning information, the bank's security management department taking corresponding security measures for the risky accounts according to the early warning information and recording the processing results.
8. The method for predicting financial account security based on multi-dimensional features according to claim 7, characterized in that: The security measures mentioned include limiting transactions and requiring identity verification.