Risk processing method and system and storage medium

Abstract

The invention discloses a risk processing method and system and a storage medium, and relates to the technical field of machine learning, and the method comprises the steps: an account opening system carries out the arrangement of collected basic information of an account opening user, and forms an account opening message; calling a risk control front interface to send the risk control request and the account opening message to a risk control system, so that the risk control system performs analysis and text similarity calculation on the account opening message based on the risk control request to obtain an analysis result and a text similarity result, and forming an account opening variable; a risk score returned by the risk control system is received through a risk control preposition, risk early warning is carried out based on the risk score, and the risk score is obtained by inputting the account opening variable into a pre-trained risk scoring model by the risk system for scoring. According to the method, risk scoring is carried out on an account opening enterprise or a user in real time through the pre-trained risk scoring model, a business party is assisted to make an account opening decision, and meanwhile, the risk identification accuracy is improved.

Description

Technical Field

[0001] The present application relates to the field of machine learning technology, and in particular to risk management methods, systems, and storage media. Background Art

[0002] With the continuous development of machine learning and machine learning technology, more and more complex algorithms are being used in anti-fraud systems. At present, stream processing technologies (such as Apache Kafka, Apache Flink, etc.) are mainly used. Anti-fraud systems can analyze transaction data in real time and make decisions on whether it is fraudulent or not within a few milliseconds. Some advanced anti-fraud systems have self-learning capabilities and can continuously adjust and optimize their models based on new data and fraud patterns.

[0003] However, the existing technology has the following disadvantages:

[0004] 1. Machine learning models rely on high-quality and diverse data sets for training. If the training data is biased, incomplete, or unrepresentative, the model may produce inaccurate results;

[0005] 2. Real-time anti-fraud methods need to make decisions in a very short time, which requires the model to be both fast and accurate;

[0006] 3. Reducing false positives and false negatives is a key challenge for anti-fraud systems. False positives can lead to customer dissatisfaction, while false negatives can lead to financial losses.

[0007] 4. Currently, corporate institutions mostly rely on expert experience and compliance review for risk review of new account opening customers. Due to the lack of sample data and the limited information dimensions available during the customer account opening stage, the application of machine learning methods is more difficult.

[0008] In summary, how to solve the above-mentioned shortcomings of the existing technologies is a major challenge facing the anti-fraud field today. Summary of the invention

[0009] The main purpose of this application is to provide a risk management method, system and storage medium, aiming to solve the technical problem of how to ensure the accuracy of risk scoring when faced with complex social relationships and multi-dimensional data.

[0010] To achieve the above objectives, the present application proposes a risk management method, which is applied to an account opening system and includes:

[0011] Organize the collected basic information of the account opening users to form an account opening message;

[0012] Calling the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system can parse the account opening message and calculate the text similarity based on the risk control request to obtain the account opening variable;

[0013] The risk control front-end receives the risk score returned by the risk control system, and performs risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring.

[0014] In one embodiment, after the step of receiving the risk score returned by the risk control system through the risk control front-end and issuing a risk warning based on the risk score, the step further includes:

[0015] The account opening data is fed back to the data platform so that the data platform can perform data mining and analysis based on the account opening data.

[0016] In addition, to achieve the above purpose, the present application also proposes a risk management method, which is applied to a risk control system and includes:

[0017] Receive the risk control request and account opening message sent by the account opening system through the risk control front-end interface, parse the account opening message and calculate the text similarity based on the risk control request to obtain the account opening variable;

[0018] The account opening variable is input into a pre-trained risk scoring model for scoring, a risk score corresponding to the account opening variable is obtained, and the risk score is returned to the account opening system through the risk control front-end interface, so that the account opening system can issue a risk warning based on the risk score.

[0019] In one embodiment, before the step of inputting the account opening variable into a pre-trained risk scoring model for scoring and obtaining the risk score corresponding to the account opening variable, the step further includes:

[0020] The risk scoring model is trained through a sample balancing strategy to obtain a pre-trained risk scoring model.

[0021] In one embodiment, the step of training the risk scoring model by using a sample balancing strategy to obtain a pre-trained risk scoring model includes:

[0022] Collecting and cleaning user credit data, wherein the user credit data includes a first type of sample points and a second type of sample points;

[0023] Extracting the first type of sample points from the user credit data, and calculating preset neighbor points of the first type of sample points using a Euclidean distance formula;

[0024] Constructing a new first-category sample point on the line connecting the first-category sample point and the preset neighbor point;

[0025] The risk scoring model is trained using the second type of sample points and the new first type of sample points to obtain a pre-trained risk scoring model.

[0026] In one embodiment, the step of inputting the account opening variable into a pre-trained risk scoring model for scoring and obtaining the risk score corresponding to the account opening variable comprises:

[0027] The account opening variables are binned to obtain segmented variables for each bin;

[0028] Calculate the eigenvalues ​​corresponding to the segmentation variables of each box;

[0029] Performing a chi-square test on each of the boxes to obtain a chi-square test result;

[0030] According to the eigenvalues ​​and the chi-square test results, adjacent bins are merged to obtain the optimal binning;

[0031] The account opening variables are processed based on the optimal binning, and the account opening variables after the optimal binning are input into a pre-trained risk scoring model for scoring, so as to obtain the risk score corresponding to the account opening variable.

[0032] In one embodiment, after the step of inputting the account opening variable into a pre-trained risk scoring model for scoring, obtaining the risk score corresponding to the account opening variable, and returning the risk score to the account opening system through the risk control front-end interface so that the account opening system can issue a risk warning based on the risk score, the method further includes:

[0033] The risk control data is fed back to the data platform so that the data platform can perform data mining and analysis based on the risk control data.

[0034] In addition, to achieve the above purpose, the present application also proposes a risk management system, the risk management system includes an account opening system, and the account opening system includes:

[0035] The sorting module is used to sort the collected basic information of the account opening user to form an account opening message;

[0036] A sending module, used to call the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable;

[0037] A receiving module, configured to receive the risk score returned by the risk control system through the risk control front-end interface, and to perform risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring;

[0038] The risk management system also includes a risk control system, which includes:

[0039] A calculation module is used to receive the risk control request and account opening message sent by the account opening system through the risk control front-end interface, and parse the account opening message and calculate the text similarity based on the risk control request to obtain the parsing result and the text similarity result;

[0040] A combination module, used for combining the parsing result and the text similarity result to form an account opening variable;

[0041] The scoring module is used to input the account opening variables into a pre-trained risk scoring model for scoring, obtain the risk score corresponding to the account opening variables, and return the risk score to the account opening system through the risk control front-end interface so that the account opening system can issue a risk warning based on the risk score.

[0042] In addition, to achieve the above objectives, the present application also proposes a risk scoring system, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the risk handling method described above.

[0043] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the risk management method described above are implemented.

[0044] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the risk handling method described above are implemented.

[0045] The embodiment of the present application provides a risk handling method, system, system and storage medium, including an account opening system collating the basic information of the account opening user collected to form an account opening message; calling the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request to obtain the account opening variable; receiving the risk score returned by the risk control system through the risk control front-end, and performing risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring. The method uses a pre-trained risk scoring model to perform risk scoring on the account opening enterprise or user in real time, assists the business party in making account opening decisions, and improves the accuracy of risk identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0048] Figure 1 A schematic diagram of a process flow provided for Embodiment 1 of the risk management method of this application;

[0049] Figure 2 A schematic diagram of the process provided for the second embodiment of the risk management method of this application;

[0050] Figure 3 A schematic diagram of the process provided for the third embodiment of the risk management method of this application;

[0051] Figure 4 A schematic diagram of sample imbalance solution provided in Example 3 of the present application;

[0052] Figure 5 A brief flowchart of the risk management method provided in Example 2 of the present application;

[0053] Figure 6 This is a schematic diagram of the module structure of the risk management system of the embodiment of the present application;

[0054] Figure 7 Schematic diagram of the system structure of the hardware operating environment involved in the risk management method in the embodiment of the present application.

[0055] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0056] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0057] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0058] The main solution of the embodiment of the present application is: the account opening system organizes the basic information of the account opening user collected to form an account opening message; calls the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable; receives the risk score returned by the risk control system through the risk control front-end, and issues a risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring.

[0059] At present, the main use of stream processing technology (such as Apache Kafka, Apache Flink, etc.) allows the anti-fraud system to analyze transaction data in real time and make a decision on whether it is fraudulent or not within a few milliseconds. Some advanced anti-fraud systems have self-learning capabilities and can continuously adjust and optimize their models based on new data and fraud patterns.

[0060] However, the existing technology has the following disadvantages:

[0061] 1. Machine learning models rely on high-quality and diverse data sets for training. If the training data is biased, incomplete, or unrepresentative, the model may produce inaccurate results;

[0062] 2. Real-time anti-fraud methods need to make decisions in a very short time, which requires the model to be both fast and accurate;

[0063] 3. Reducing false positives and false negatives is a key challenge for anti-fraud systems. False positives can lead to customer dissatisfaction, while false negatives can lead to financial losses.

[0064] 4. Currently, corporate institutions mostly rely on expert experience and compliance review for risk review of new account opening customers. Due to the lack of sample data and the limited information dimensions available during the customer account opening stage, the application of machine learning methods is more difficult.

[0065] The present application provides a solution, which organizes the basic information of the account opening user collected through the account opening system to form an account opening message; calls the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system can parse the account opening message and calculate the text similarity based on the risk control request to obtain the account opening variable; receives the risk score returned by the risk control system through the risk control front-end, and performs risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring. This solution uses a pre-trained risk scoring model to perform risk scoring on the account opening enterprise or user in real time, assists the business party in making account opening decisions, and improves the accuracy of risk identification. At the same time, this solution is different from the traditional sampling solution, and uses "artificial" samples to solve the problem of sample imbalance, thereby improving the generalization ability of the risk scoring model; by combining the text similarity calculation method, the calculation of massive message data can be completed in milliseconds, thereby improving the accuracy of the risk scoring model. Use machine learning methods to identify account opening risks and supplement expert experience.

[0066] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, etc. The following takes a personal computer as an example to illustrate this embodiment and the following embodiments.

[0067] Based on this, the present application embodiment provides a risk management method, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the risk management method of the present application.

[0068] In this embodiment, the risk handling method is applied to an account opening system, and the method includes steps S10 to S30:

[0069] Step S10, collating the collected basic information of the account opening user to form an account opening message;

[0070] At present, institutions in related fields mainly use expert rules to conduct risk detection in corporate account opening risk control. Faced with increasingly complex social networks and massive amounts of data, their shortcomings in accuracy and risk prevention and control adaptability are becoming increasingly obvious.

[0071] Based on this, it is proposed to use machine learning methods to mine and calculate the multi-dimensional data of account-opening enterprises to predict their risks in real time and supplement the existing risk identification methods.

[0072] It should be noted that the basic information of the account opening user includes but is not limited to the basic information and preliminary review information of the account opening user or enterprise.

[0073] Optionally, first, the collected basic information of account opening users or enterprises and preliminary review information are organized in a certain structured manner, and a predefined template is used to create an account opening message, which contains the basic information of all account opening users and the opinions and results of the preliminary review.

[0074] Step S20, calling the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable;

[0075] It should be noted that the risk control request includes but is not limited to request identifier, timestamp, request type and other information, which is used to identify the purpose of the current risk control check.

[0076] Optionally, after obtaining the account opening message, the account opening system sends a risk control request and the account opening message to the risk control system by calling the risk control front-end interface, and the risk control system parses the account opening message and calculates text similarity in response to the risk control request.

[0077] Specifically, the process of the risk control system parsing the account opening message and calculating the text similarity is as follows: the risk control system verifies whether the message conforms to the expected data format. If the format is incorrect, the risk control system may directly return an error message. Once the format verification is passed, the risk control system will extract all relevant field information from the account opening message, such as company name, legal representative information, registered address, etc. Subsequently, the text similarity of the account opening message is calculated using the TF-IDF (TermFrequency-Inverse Document Frequency) method built into ES (Elasticsearch). Finally, all the relevant field information extracted and the text similarity results of the account opening message are organized into the input variables required by the pre-trained risk scoring model. In this embodiment, the account opening variable is used as the input variable.

[0078] Step S30, receiving the risk score returned by the risk control system through the risk control front-end interface, and performing risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring.

[0079] Optionally, after the risk control system obtains the account opening variable, the model interface is called to input the account opening variable into a pre-trained risk scoring model. After the pre-trained risk scoring model outputs a risk score based on the account opening variable, the risk control system returns it to the account opening system through the risk control front-end interface. After the account opening system receives the risk score, the risk score is a value between 0 and 1, indicating the risk level of the account opening request. The higher the score, the greater the risk.

[0080] Set different risk thresholds based on business needs and historical data analysis. For example:

[0081] Low risk: score < 0.3

[0082] Medium risk: 0.3≤score<0.7

[0083] High risk: score ≥ 0.7

[0084] After setting the risk threshold, the account opening system starts the real-time monitoring function. When a new account opening request arrives, its risk score is immediately calculated through the risk control system. If the calculated risk score exceeds the preset threshold, the account opening system automatically triggers a risk warning.

[0085] It should be noted that, in this embodiment, the pre-trained risk scoring model is obtained by training the risk scoring model based on a preset sample balance strategy, which can solve the problem of sample imbalance and thus improve the generalization ability of the risk scoring model.

[0086] In addition, it should be noted that in the pre-trained risk scoring model, the account opening variable is binned and converted into a finite number of intervals (or boxes), thereby simplifying the data and improving the interpretability and stability of the model.

[0087] After completing the risk assessment and early warning of the account opening system, the account opening system will return the account opening data of the entire execution process to the data platform, and continuously iterate and upgrade the entire execution process by mining and analyzing the account opening data.

[0088] Through the above-mentioned embodiment method, the account opening system organizes the basic information of the account opening user collected to form an account opening message; calls the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable; receives the risk score returned by the risk control system through the risk control front-end, and performs risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring. This method uses a pre-trained risk scoring model to perform risk scoring on the account opening enterprise or user in real time, assists the business party in making account opening decisions, and improves the accuracy of risk identification.

[0089] Based on this, the present application embodiment provides a risk management method, referring to Figure 2 , Figure 2 This is a flow chart of the second embodiment of the risk management method of the present application.

[0090] In this embodiment, the risk handling method is applied to a risk control system, and the method includes steps S40 to S60:

[0091] Step S40, receiving the risk control request and the account opening message sent by the account opening system through the risk control front-end interface, and parsing the account opening message and calculating the text similarity based on the risk control request to obtain the parsing result and the text similarity result;

[0092] It should be noted that text similarity calculation methods include but are not limited to TF-IDF (Term Frequency-Inverse Document Frequency) weighting, cosine similarity, edit distance, etc. This embodiment takes the TF-IDF similarity calculation method as an example for illustration.

[0093] Optionally, the risk control pre-processing is to evaluate the potential risks through a series of means before formal account opening or trading. In this embodiment, it includes but is not limited to parsing the received message, querying the database (such as ES library), and using TF-IDF and other methods to calculate the text similarity to prepare the variables required by the pre-trained risk scoring model, specifically:

[0094] First, in order to extract useful information from the account opening message sent by the account opening system, the risk control system uses the XML / JSON parsing library to parse the structured data message. For unstructured text data, regular expressions or natural language processing tools are used to extract key information to ensure that all necessary fields are correctly identified and converted into a form suitable for further analysis. Among them, the useful information extracted from the account opening message includes but is not limited to the business information of the account opening enterprise, multiplication information, transaction behavior, social relations, administrative records, text mining, registration behavior and other information.

[0095] In order to quickly retrieve historical records or other relevant information and assist the risk control system in the risk decision-making process, this embodiment defines appropriate indexes and document types according to business needs through the risk control system, builds efficient query statements, and queries related attributes or cases in the ES library according to the query statements. For example, the customer's past behavior records are searched in the ES library by customer ID.

[0096] In order to measure the correlation between different message texts and identify abnormal situations such as fraud, this embodiment uses the risk control system to calculate the frequency of each term in the account opening message in the entire document collection and its inverse document frequency. And compare the content of the risk control request received by the risk control system with the known risk samples. At the same time, each document is represented in vector form, where each dimension corresponds to a term, and its value is determined by the TF-IDF weight of the term. Subsequently, the cosine similarity method is used to calculate the distance between the two documents to determine the degree of similarity between them. This embodiment combines the TF-IDF method built into ES to achieve millisecond-level calculation of the similarity of massive texts.

[0097] Through the above steps, by traversing the ES library query and offline combination calculation, the business information, multiplication information, transaction behavior, social relations, administrative records, text mining, registration behavior and other data of the account-opening enterprise are integrated, and the multidimensional model variables are calculated quickly in real time, which can effectively improve the accuracy and response speed of the risk system and reduce the misjudgment rate.

[0098] Step S50, combining the parsing result and the text similarity result to form an account opening variable;

[0099] Optionally, after completing the above processing, the parsing results and text similarity results corresponding to the account opening message are combined to form an account opening variable, and the account opening variable is used as the final input variable of the pre-trained risk scoring model.

[0100] Step S60, input the account opening variable into a pre-trained risk scoring model for scoring, obtain the risk score corresponding to the account opening variable, and return the risk score to the account opening system through the risk control front-end interface, so that the account opening system can issue a risk warning based on the risk score.

[0101] It should be noted that the risk score is a value between 0 and 1, indicating the risk level of the account opening request. The higher the risk score, the greater the fraud risk of the user or enterprise.

[0102] It is worth noting that in order to prevent high-risk customers from entering the system and reduce the occurrence of fraud, step S60 is executed to perform risk scoring on account opening users or enterprises in real time, and it is impossible to assist in account opening decisions, thereby improving the risk scoring rate.

[0103] In a feasible implementation, step S60 may include steps S61 to S65:

[0104] Step S61, binning the account opening variables to obtain segmented variables for each bin;

[0105] It should be noted that in order to enable the pre-trained risk scoring model to better identify potential risk factors in the account opening variables, this embodiment needs to perform data processing on the account opening variables before the risk scoring model performs the risk scoring process, that is, use a greedy algorithm combined with a chi-square test to optimally bin the account opening variables, which can improve the interpretability and predictive power of the risk scoring model.

[0106] Among them, the greedy algorithm merges adjacent intervals with the least overall information loss during the binning process until the number of bins reaches a preset number.

[0107] The chi-square test is used to detect whether there is a significant association between two variables. In the binning process, the chi-square statistic under the binning scheme is compared to determine which method can better maintain or reveal the correlation between the original variable and the target variable.

[0108] Optionally, the account opening variables are binned in the pre-trained risk scoring model, that is, the account opening variables are segmented within a predetermined numerical range to make the distribution of good and bad in each segment more obvious, and to deeply explain the different variations of good and bad enterprises and users at the data level to improve the distinguishing effect of the risk scoring model.

[0109] Step S62, calculating the characteristic value corresponding to the segmentation variable of each box;

[0110] It should be noted that the characteristic values ​​include but are not limited to IV (Information Value) and WOE (Weight Of Evidence), where IV is used to evaluate the predictive ability of variables, and WOE is a coding method that can convert categorical variables into numerical form.

[0111] Optionally, after obtaining the segmentation variables in each box through step S51, for each box, the corresponding WOE value is calculated. The WOE value reflects the difference between the good and bad sample ratio in the box and the good and bad sample ratio of the whole, which can be obtained by the following calculation formula:

[0112] WOE = In (number of good samples / number of bad samples)

[0113] Then, for each box, its contribution to the total IV value is calculated based on the calculated WOE value. This can be obtained by the following calculation formula:

[0114] IV = (number of good samples - number of bad samples) × WOE

[0115] In the above steps, by calculating the IV value and WOE weight corresponding to the segmentation variables of each box, we can effectively evaluate the importance of variables, optimize the binning scheme, generate high-quality features, improve model performance, and provide users with better business understanding and explanation.

[0116] Step S63, performing a chi-square test on each of the boxes to obtain a chi-square test result;

[0117] It should be noted that after binning, it is assumed that there is some correlation between adjacent boxes. For example, two boxes of a vector have similar distributions of good and bad samples, so the two adjacent boxes can be merged. In order to verify this hypothesis, in this embodiment, a chi-square test is performed on each box by calculating the chi-square confidence PV (p-value) value of each box to evaluate whether there is a significant difference in the distribution of good and bad samples between adjacent boxes. The smaller the p-value, the greater the difference between the two boxes, that is, the two boxes are statistically significantly different.

[0118] Optionally, first for each pair of adjacent boxes, build a 2x2 contingency table to count the number of good samples and bad samples. Then, calculate the expected frequency based on the number of good samples and bad samples, and the chi-square statistic. Calculate the PV value based on the chi-square statistic using the chi-square distribution table.

[0119] Step S64, merging adjacent bins according to the eigenvalues ​​and the chi-square test results to obtain optimal binning;

[0120] Optionally, a preset threshold (such as 0.05) corresponding to the PV value is set. If the PV value between adjacent boxes is greater than the preset threshold (such as 0.05), it is considered that the two boxes have no significant difference and can be merged. Repeat the above process until the PV value between all adjacent boxes is less than the preset threshold, or the number of merged boxes reaches the minimum box number limit.

[0121] Finally, an optimized binning scheme is obtained so that each box is statistically significantly different, and each box has good predictive ability through IV value and WOE weight.

[0122] Step S65, processing the account opening variables based on the optimal binning, and inputting the account opening variables after the optimal binning into a pre-trained risk scoring model for scoring, so as to obtain the risk score corresponding to the account opening variable.

[0123] Optionally, the optimized binning scheme is applied to the account opening variable, that is, the original account opening variable is replaced with its corresponding WOE value or the new binning is directly used as the input feature. Then, the account opening variable after the optimal binning is called the model interface and input into the pre-trained risk scoring model for scoring to obtain the risk score corresponding to the account opening variable.

[0124] Finally, the pre-trained risk scoring model outputs the risk score of each account opening user's application, and the risk control system returns the risk score to the account opening system through the risk control front-end interface. After receiving the risk score, the account opening system compares it with the pre-set risk threshold to determine whether to trigger the risk warning function.

[0125] In addition, after the risk control system returns the risk score to the account opening system through the risk control front-end interface, it also flows the risk control data of the entire execution process back to the data platform. By mining and analyzing the account opening data, the entire execution process is continuously iterated and upgraded.

[0126] Through the above-mentioned embodiment method, the risk control system receives the risk control request and account opening message sent by the account opening system through the risk control front-end interface, and parses the account opening message based on the risk control request and calculates the text similarity to obtain the parsing result and the text similarity result; the parsing result and the text similarity result are combined to form an account opening variable; the account opening variable is input into a pre-trained risk scoring model for scoring, and the risk score corresponding to the account opening variable is obtained, and the risk score is returned to the account opening system through the risk control front-end interface, so that the account opening system can provide risk warning based on the risk score. This method uses a pre-trained risk scoring model to score the risk of the account opening enterprise or user in real time, assists the business party in making account opening decisions, and improves the accuracy of risk identification. And combined with the text similarity calculation method, the calculation of massive message data can be completed in milliseconds, improving the accuracy of the risk scoring model. Use machine learning methods to identify account opening risks and supplement expert experience.

[0127] Based on the second embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the above-mentioned second embodiment can refer to the above introduction, and will not be repeated in the following. Figure 3 Before step S60, the risk handling method further includes step S601:

[0128] Step S501, training the risk scoring model through a sample balancing strategy to obtain a pre-trained risk scoring model.

[0129] Optionally, in order to more fully characterize the potential risks of users or enterprises, this embodiment uses a multi-dimensional user credit data source as a training set for the risk scoring model. The user credit data includes but is not limited to corporate business information, member information, and administrative records. In particular, the model input variables constructed by mining the transaction behavior, social associations, and text similarity of the enterprise play an effective role in improving the effect of the risk scoring model.

[0130] Furthermore, different from the traditional sampling method, this embodiment uses "artificial" samples to solve the problem of sample imbalance and improve the accuracy and generalization ability of the risk scoring model.

[0131] Please refer to Figure 4 , Figure 4 A schematic diagram of sample imbalance solution provided by the risk handling method of this embodiment. Figure 4It can be seen that, assuming that the yellow dots in the figure represent the first-class sample points, and the first-class sample points represent the minority class sample points, the green dots represent the new first-class sample points, that is, the artificially synthesized minority class sample points, and the blue dots represent the second-class sample points, and the second-class sample points represent the majority class samples. First, randomly select a minority class sample point, such as point A, B or C in the figure, use the Euclidean distance to calculate the K nearest neighbor points of the minority class sample point, assuming K = 3, and randomly select one of the neighbor points (such as point D). Then randomly construct a new first-class sample point on the line connecting point A and point D, and use the new first-class sample point as the artificially synthesized minority class sample point. Repeat the above steps until enough minority class samples are generated. Finally, filter the generated samples and eliminate the noise data, such as the green point constructed by point C in the figure.

[0132] Finally, the risk scoring model is trained using majority class sample points, artificially synthesized minority class sample points, and originally existing minority class sample points to obtain a pre-trained risk scoring model.

[0133] Through the above-mentioned embodiment method, a classifier is first constructed between the existing good and bad samples, and then the minority class samples are randomly sampled on the minority class sample side of the classifier to create minority class samples. This method not only solves the sample imbalance problem, but also improves the generalization ability of the risk scoring model.

[0134] For example, to help understand the implementation process of the risk handling method obtained by combining the above-mentioned embodiment 1 and embodiment 2, please refer to Figure 5 , Figure 5 A brief flow chart of the risk treatment method is provided, specifically:

[0135] The starting point of this embodiment is to conduct risk scoring for account-opening enterprises or users in real time, assist the business party in making account opening decisions, and at the same time improve the risk identification rate.

[0136] Optionally, first, the risk scoring model is developed and deployed, and the process mainly includes data collection, data cleaning, feature engineering, model training, online model deployment, model interface monitoring, and model iteration optimization. This embodiment expands the data scope considered in the conventional account opening review, and deeply mines the social relationships, transaction behaviors, and corporate registration naming habits of the account opening enterprise. The main steps are: cleaning the collected multi-dimensional user credit data source, and solving the problem of imbalanced training samples through a preset balanced sample strategy, thereby improving the accuracy and generalization ability of the model. Among them, in this embodiment, the preset balanced sample strategy is mainly to create a few-class sample by first building a classifier between the existing good and bad samples, and then randomly sampling on the few-class sample side of the classifier. Subsequently, the greedy algorithm and the chi-square test method are combined to optimally bin the account opening variables, effectively optimize the variable binning, and improve the performance and stability of the model.

[0137] When the risk scoring model is deployed and developed, the account opening system will organize the basic information and preliminary review information of the enterprise that needs to open an account into an account opening message, and call the risk control front-end interface to send the account opening message to the risk control system. The risk control system parses the account opening message, queries the ES library, calculates the TF-IDF text similarity, and organizes the account opening variables required by the risk scoring model, and calls the model interface to input the account opening variables into the risk scoring model, and the risk scoring model outputs the risk score. The risk control system then returns the risk score to the account opening system through the risk control front-end interface. The entire process is completed in milliseconds, realizing risk warning for the account opening enterprise.

[0138] Finally, the account opening system and risk control system will return the log data of the entire process to the data platform, and continuously iterate and upgrade the entire process by mining and analyzing the log data.

[0139] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the risk management method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0140] This application also provides a risk management system, please refer to Figure 6 , the risk management system includes an account opening system, and the account opening system includes:

[0141] The sorting module 10 is used to sort the collected basic information of the account opening user to form an account opening message;

[0142] The sending module 20 is used to call the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable;

[0143] A receiving module 30 is used to receive the risk score returned by the risk control system through the risk control front-end interface, and to perform risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring;

[0144] The risk management system also includes a risk control system, which includes:

[0145] The calculation module 40 is used to receive the risk control request and the account opening message sent by the account opening system through the risk control front-end interface, and parse the account opening message and calculate the text similarity based on the risk control request to obtain the parsing result and the text similarity result;

[0146] A combination module 50, used for combining the parsing result and the text similarity result to form an account opening variable;

[0147] The scoring module 60 is used to input the account opening variables into a pre-trained risk scoring model for scoring, obtain the risk score corresponding to the account opening variables, and return the risk score to the account opening system through the risk control front-end interface so that the account opening system can issue a risk warning based on the risk score.

[0148] The risk management system provided by the present application adopts the risk management method in the above embodiment, which can solve the technical problem of how to ensure the accuracy of risk scoring when facing complex social relationships and multidimensional data. Compared with the prior art, the beneficial effects of the risk management system provided by the present application are the same as the beneficial effects of the risk management method provided by the above embodiment, and the other technical features of the risk management system are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0149] The present application provides a risk scoring system, which includes: at least one processor; and a memory in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the risk handling method in the above-mentioned embodiment 1.

[0150] Reference below Figure 7, which shows a schematic diagram of the structure of the risk scoring system suitable for implementing the embodiment of the present application. The risk scoring system in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The risk scoring system shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0151] like Figure 7 As shown, the risk scoring system may include a processing system 1001 (e.g., a central processing unit, a graphics processor, etc.), which may perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage system 1003 to a random access memory (RAM) 1004. In RAM 1004, various programs and data required for the operation of the risk scoring system are also stored. The processing system 1001, ROM 1002, and RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems may be connected to the I / O interface 1006: an input system 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output system 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage system 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication system 1009. The communication system 1009 can allow the risk scoring system to communicate wirelessly or wired with other systems to exchange data. Although the risk scoring system with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0152] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication system, or installed from a storage system 1003, or installed from a ROM 1002. When the computer program is executed by the processing system 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0153] The risk scoring system provided by this application adopts the risk handling method in the above embodiment, which can solve the technical problem of how to ensure the accuracy of fraud risk identification when facing complex social relationships and multi-dimensional data. Compared with the prior art, the beneficial effects of the risk scoring system provided by this application are the same as the beneficial effects of the risk handling method provided by the above embodiment, and the other technical features of the risk scoring system are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0154] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0155] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0156] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, wherein the computer-readable program instructions are used to execute the risk handling method in the above-mentioned embodiment.

[0157] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: 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. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0158] The computer-readable storage medium may be included in the risk scoring system; or may exist independently without being incorporated into the risk scoring system.

[0159] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the risk scoring system, the risk scoring system: the account opening system organizes the basic information of the account opening user collected to form an account opening message; calls the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable; receives the risk score returned by the risk control system through the risk control front-end, and issues a risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring.

[0160] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0161] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0162] The modules involved in the embodiments of the present application may be implemented by software or hardware, wherein the name of the module does not limit the unit itself in some cases.

[0163] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned risk processing method, and can solve the technical problem of how to ensure the accuracy of risk scoring when facing complex social relationships and multidimensional data. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the risk processing method provided in the above-mentioned embodiment, and will not be repeated here.

[0164] The present application also provides a computer program product, including a computer program, which implements the steps of the above-mentioned risk handling method when executed by a processor.

[0165] The computer program product provided by this application can solve the technical problem of how to ensure the accuracy of risk scoring when facing complex social relationships and multidimensional data. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the risk handling method provided by the above embodiment, which will not be repeated here.

[0166] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A risk management method, characterized in that: The method is applied to an account opening system, and the method comprises: Organize the collected basic information of the account opening users to form an account opening message; Calling the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable; The risk score returned by the risk control system is received through the risk control front-end interface, and a risk warning is performed based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring.

2. The method according to claim 1, characterized in that After the step of receiving the risk score returned by the risk control system through the risk control front-end interface and performing risk warning based on the risk score, the method further includes: The account opening data is fed back to the data platform so that the data platform can perform data mining and analysis based on the account opening data.

3. A risk management method, characterized in that: The method is applied to a wind control system, and the method comprises: Receiving a risk control request and an account opening message sent by the account opening system through a risk control front-end interface, parsing the account opening message and calculating text similarity based on the risk control request, and obtaining a parsing result and a text similarity result; Combining the parsing result and the text similarity result to form an account opening variable; The account opening variable is input into a pre-trained risk scoring model for scoring, a risk score corresponding to the account opening variable is obtained, and the risk score is returned to the account opening system through the risk control front-end interface, so that the account opening system can issue a risk warning based on the risk score.

4. The method according to claim 3, characterized in that Before the step of inputting the account opening variable into a pre-trained risk scoring model for scoring and obtaining the risk score corresponding to the account opening variable, the step further includes: The risk scoring model is trained through a sample balancing strategy to obtain a pre-trained risk scoring model.

5. The method according to claim 4, characterized in that The step of training the risk scoring model by using the sample balancing strategy to obtain a pre-trained risk scoring model includes: Collecting and cleaning user credit data, wherein the user credit data includes a first type of sample points and a second type of sample points; Extracting the first type of sample points from the user credit data, and calculating preset neighbor points of the first type of sample points using a Euclidean distance formula; Constructing a new first-category sample point on the line connecting the first-category sample point and the preset neighbor point; The risk scoring model is trained using the second type of sample points and the first type of sample points to obtain a pre-trained risk scoring model.

6. The method according to claim 3, characterized in that The step of inputting the account opening variable into a pre-trained risk scoring model for scoring and obtaining the risk score corresponding to the account opening variable comprises: The account opening variables are binned to obtain segmented variables for each bin; Calculate the eigenvalues ​​corresponding to the segmentation variables of each box; Performing a chi-square test on each of the boxes to obtain a chi-square test result; According to the eigenvalues ​​and the chi-square test results, adjacent bins are merged to obtain the optimal binning; The account opening variables are processed based on the optimal binning, and the account opening variables after the optimal binning are input into a pre-trained risk scoring model for scoring, so as to obtain the risk score corresponding to the account opening variable.

7. The method according to claim 3, characterized in that After the step of inputting the account opening variable into a pre-trained risk scoring model for scoring, obtaining the risk score corresponding to the account opening variable, and returning the risk score to the account opening system through the risk control front-end interface so that the account opening system can issue a risk warning based on the risk score, the method further includes: The risk control data is fed back to the data platform so that the data platform can perform data mining and analysis based on the risk control data.

8. A risk management system, characterized in that: The risk handling system includes an account opening system, and the account opening system includes: The sorting module is used to sort the collected basic information of the account opening user to form an account opening message; A sending module, used to call the risk control front-end interface to send the risk control request and the account opening message to the risk control system, so that the risk control system parses the account opening message and calculates the text similarity based on the risk control request, obtains the parsing result and the text similarity result, and combines the parsing result and the text similarity result to form an account opening variable; A receiving module, configured to receive the risk score returned by the risk control system through the risk control front-end interface, and to perform risk warning based on the risk score, wherein the risk score is obtained by the risk system inputting the account opening variable into a pre-trained risk scoring model for scoring; The risk management system also includes a risk control system, which includes: A calculation module is used to receive the risk control request and account opening message sent by the account opening system through the risk control front-end interface, and parse the account opening message and calculate the text similarity based on the risk control request to obtain the parsing result and the text similarity result; A combination module, used for combining the parsing result and the text similarity result to form an account opening variable; The scoring module is used to input the account opening variables into a pre-trained risk scoring model for scoring, obtain the risk score corresponding to the account opening variables, and return the risk score to the account opening system through the risk control front-end interface so that the account opening system can issue risk warnings based on the risk score.

9. A risk management system, characterized in that: The system comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the risk handling method according to any one of claims 1 to 2 or claims 3 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the risk management method according to any one of claims 1 to 2 or claims 3 to 7 are implemented.

Images