An intelligent early warning system for customer potential risks based on SAAS technology

Abstract

The present invention belongs to the technical field of risk early warning management, and specifically discloses a customer potential risk intelligent early warning system based on SAAS technology, which includes: transaction data collection module, transaction data analysis module, early warning trigger assessment module, early warning trigger confirmation module and risk early warning feedback terminal. The present invention collects the tracking transaction data set of the target customer and the current transaction data to be confirmed, analyzes and obtains transaction feature information, effectively solves the current problem of insufficient depth of customer analysis, avoids the current lack of personalized consideration, improves the accuracy and authenticity of customer transaction early warning assessment, and evaluates whether the early warning is triggered by combining the customer's transaction feature information, and confirms the early warning risk category, which facilitates subsequent precise and targeted early warnings, avoids the shortcomings of the current general and relatively universal early warning methods, and improves the effectiveness and timeliness of the early warning.

Description

Technical Field

[0001] The present invention belongs to the technical field of risk early warning management, and in particular, relates to an intelligent early warning system for potential customer risks based on SAAS technology. Background Art

[0002] With the intensification of market competition and the increase in uncertainty in the economic environment, enterprises are facing many potential customer risks, such as credit risk, market risk, operational risk, etc. In order to effectively manage these risks, the importance of timely identification and early warning of potential customer risks is highlighted, especially for special enterprises such as banks.

[0003] Prior art, such as a Chinese invention patent application with authorization announcement number CN109272396B, discloses a customer risk warning method, device, computer equipment and storage medium based on big data analysis, wherein the method comprises: obtaining basic risk data of target customers, wherein the basic risk data comprises an industry identifier. Extracting basic risk indicators of the basic risk data. Obtaining associated risk data of the target customers. Determining the associated risk indicators corresponding to the target customers based on the associated risk data. Obtaining a corresponding risk prediction model based on the industry identifier. Inputting the basic risk indicators and the associated risk indicators into the risk prediction model, outputting a corresponding risk score, and performing customer risk warning based on the risk score. This method can improve the accuracy of risk warnings.

[0004] Another example of the prior art is a customer risk data monitoring method based on in-bank data and external data disclosed in a Chinese invention patent application with authorization announcement number CN111913994B, which includes a system source that penetrates the data into the system based on in-bank business data provided by a big data platform, the credit information of the People's Bank of China, and the data content of a third-party data supplier. The system mart layer uses the subject domain division method based on the definition of business data, and also filters and integrates the data through filtering and calculation of some metering scripts. And it is combined and calculated based on the rule engine of the model layer. The model layer performs metering analysis based on the mart data through the fico early warning rule engine, identifies risks, generates early warning results, and merges them into a task report through the workflow, which is a complete set of processes that is pushed to managers.

[0005] Regarding the above two technical solutions, it is obvious that the current customer potential risk warning still has the following deficiencies: 1. Insufficient depth of customer analysis and lack of personalized consideration. The current warning method is a general one and is relatively universal. It does not combine the customer's transaction characteristics and other rules for detailed assessment, and cannot guarantee the timeliness and accuracy of the warning.

[0006] 2. Insufficient control over changes in the customer's time dimension results in a certain lack of effectiveness and pertinence in early warnings. At the same time, the impact of customer fluctuations is ignored, which leads to a certain deviation in the accuracy of risk prediction. Summary of the invention

[0007] In view of this, in order to solve the problems raised in the above background technology, a customer potential risk intelligent early warning system based on SAAS technology is proposed.

[0008] The purpose of the present invention can be achieved through the following technical solutions: The present invention provides a customer potential risk intelligent early warning system based on SAAS technology, the system comprising: a transaction data collection module, used to record the current transaction customer as the target customer, and collect the target customer's tracking transaction data and the current transaction data to be confirmed from the cloud.

[0009] The transaction data analysis module is used to analyze the tracking transaction data of target customers and obtain the transaction feature information of target users.

[0010] The warning trigger assessment module is used to assess whether the warning is triggered based on the current transaction data to be confirmed and the transaction feature information. If not, the warning instruction is not started. If so, the warning trigger confirmation module is started.

[0011] The early warning trigger confirmation module is used to confirm the early warning risk category of the target customer.

[0012] The risk warning feedback terminal is used to send corresponding warning instructions to the target customer's transaction web interface based on the target customer's warning risk category.

[0013] Compared with the prior art, the present invention has the following beneficial effects: (1) By collecting the target customer's tracking transaction data set and the current transaction data to be confirmed, and analyzing the transaction feature information, the present invention effectively solves the current problem of insufficient depth of customer analysis, avoids the current lack of personalized considerations, and improves the accuracy and authenticity of customer transaction warning assessment. At the same time, it also facilitates the pertinence and processing efficiency of subsequent customer transaction warning processing, and also improves the effectiveness of the warning.

[0014] (2) The present invention evaluates whether the warning is triggered by combining the customer's transaction feature information and confirms the warning risk category, which facilitates subsequent precise and targeted warnings, avoids the shortcomings of the current general and relatively common warning methods, fully considers the customer's transaction characteristics and other laws, realizes the customer's detailed warning evaluation, and ensures the timeliness and accuracy of subsequent transaction warnings.

[0015] (3) The present invention constructs a transaction amount curve and sets conditions according to the slope and fluctuation points to confirm the transaction trend type of the target customer. It fully combines the development impact and fluctuation impact of the customer's transaction, provides a data basis for the subsequent setting of the safe excess tendency transaction amount value, thereby improving the reference value of the evaluation results of whether the early warning is triggered or not, and also provides assistance for customers to formulate more targeted management strategies.

[0016] (4) The present invention obtains a comprehensive fixity index by comprehensively considering the fixity of the three dimensions of transaction time, transaction region and transaction account, and setting appropriate weights according to business needs. This helps to more comprehensively understand the customer's transaction behavior pattern, solves the current problem of insufficient control over changes in the customer's time dimension, and further improves the accuracy of subsequent early warning trigger assessment.

[0017] (5) When evaluating whether an early warning is triggered, the present invention makes a comprehensive consideration of the three dimensions of amount, payment area, and payment account, and sets a transaction amount value with a safety excess tendency and a confirmation payment safety level. This allows for a more comprehensive assessment of transaction risks and reduces security loopholes that may be missed in a single dimension. It also helps reduce false positives and missed negatives, improves the overall reliability, sensitivity, and accuracy of early warnings, and thereby improves customer transaction security. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0019] Figure 1 It is a schematic diagram of the connection of the system modules of the present invention.

[0020] Figure 2 It is a schematic flow chart of the overall implementation steps of the present invention.

[0021] Figure 3 It is a flow chart of the early warning triggering assessment module of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0023] See also Figure 1 and Figure 2 As shown, the present invention provides a customer potential risk intelligent early warning system based on SAAS technology, the system includes: a transaction data collection module, a transaction data analysis module, an early warning trigger assessment module, an early warning trigger confirmation module and a risk early warning feedback terminal.

[0024] In the above, the transaction data analysis module is connected to the transaction data collection module and the early warning trigger assessment module respectively, the early warning trigger confirmation module is connected to the early warning trigger assessment module and the risk early warning feedback terminal respectively, and the early warning trigger assessment module is also connected to the transaction data collection module.

[0025] The transaction data collection module is used to record the current transaction customer as the target customer, and collect the target customer's tracking transaction data and the current transaction data to be confirmed from the cloud.

[0026] Specifically, the tracking transaction data includes the cumulative number of transactions and the transaction amount, transaction date, payment area and payment account of each cumulative transaction.

[0027] In yet another specific aspect, the current transaction data to be confirmed includes the transaction amount to be confirmed, the payment region and payment account number of the transaction amount to be confirmed.

[0028] The transaction data analysis module is used to analyze the tracking transaction data of target customers to obtain transaction feature information of target users.

[0029] Specifically, analyzing the tracking transaction data of the target customer includes: A1, extracting the cumulative number of transactions and the transaction amount, transaction date, payment area and payment account of each cumulative transaction from the tracking transaction data.

[0030] A2. Confirm the target customers' trading trend types, including growth, volatility and stability.

[0031] A3. Filter out the highest transaction amount and the lowest transaction amount from the accumulated transaction amounts, and construct a transaction amount range.

[0032] A4. Select the transaction amount interval division method and the number of intervals, divide the transaction amount interval into transaction amount sub-intervals, count the cumulative number of transactions in each transaction amount sub-interval, and confirm the target customer's preferred transaction amount interval.

[0033] A5. Based on the payment area and payment account, analyze the transaction fixity of the target customer, recorded as .

[0034] A6. The transaction trend type, the highest transaction amount, the preferred transaction amount range and the transaction fixity are used as the transaction characteristic information of the target user.

[0035] The embodiment of the present invention collects the target customer's tracking transaction data set and the current transaction data to be confirmed, and analyzes the transaction feature information, thereby effectively solving the current problem of insufficient depth of customer analysis, avoiding the current lack of personalized considerations, and improving the accuracy and authenticity of customer transaction warning evaluation. At the same time, it also facilitates the targetedness and processing efficiency of subsequent customer transaction warning processing, and also improves the effectiveness of the warning.

[0036] Furthermore, regarding the confirmation of the transaction trend type of the target customer described in step A2, it includes: A21, constructing a transaction amount curve with the cumulative transaction order as the horizontal axis and the transaction amount as the vertical axis, and extracting the slope therefrom as the transaction growth rate, and at the same time extracting the number of fluctuation points and comparing it with the recorded cumulative number of transactions, and taking the ratio as the fluctuation ratio.

[0037] A22. If the transaction growth rate is greater than or equal to the set reference transaction growth rate as condition 1, the volatility ratio is less than or equal to the set volatility ratio as condition 2.

[0038] A23. If conditions 1 and 2 are met, growth will be used as the trading trend type. If only condition 2 is met, stability will be used as the trading trend type. If only condition 1 is met, volatility will be used as the trading trend type.

[0039] It should be added that, in a specific embodiment, the reference transaction growth rate can be set to a real number greater than 0. For example, the specific value can be 0.1. The specific value of the volatility ratio is mainly set according to the cumulative number of transactions recorded. For example, if the cumulative number of transactions recorded is 100 times, the volatility ratio can be 0.3.

[0040] The embodiment of the present invention constructs a transaction amount curve and sets conditions according to the slope and fluctuation points, thereby confirming the transaction trend type of the target customer. It fully combines the development impact and fluctuation impact of the customer's transaction, provides a data basis for the subsequent setting of the transaction amount value that safely exceeds the tendency, thereby improving the reference value of the evaluation results of whether the early warning is triggered or not, and also provides assistance for customers to formulate more targeted management strategies.

[0041] Furthermore, the specific selection process for selecting the transaction amount interval division method in step A4 is as follows: calculate the standard deviation of the cumulative transaction amounts of each time as the transaction amount difference. If the transaction amount difference is less than or equal to the set reference transaction amount difference, equal-width interval division is used as the transaction amount interval division method.

[0042] If the difference in transaction amounts is greater than the set reference transaction amount difference, unequal width intervals will be used as the transaction interval division method.

[0043] In a specific embodiment, the value range of the reference transaction amount difference is usually determined according to the specific business scenario and data distribution. Assuming the scenario where the transaction amount is relatively stable and has small fluctuations, such as small transactions, daily consumption, etc., the reference transaction amount difference can be set to 100. Assuming the scenario where the transaction amount fluctuates but is generally controllable, such as medium-sized transactions, regular transactions between enterprises, etc., the reference transaction amount difference can be set to 500. Assuming the scenario where the transaction amount fluctuates greatly and is difficult to predict, such as financial market transactions, stock transactions, etc., the reference transaction amount difference can be set to 1000.

[0044] Furthermore, the specific process of selecting the number of intervals in step A4 is as follows: Extract the transaction amount difference, recorded as .

[0045] Calculate the average transaction amount of each cumulative transaction to get the average transaction amount, and record the difference between the highest transaction amount and the average transaction amount as , set interval setting demand factor , , is a natural constant, To set the reference transaction amount difference, The transaction amount difference is used to set the reference.

[0046] Calculate the number of partition intervals , , The number of intervals divided by the reference benchmark, The floor symbol.

[0047] It should be added that the interval setting demand factor is set according to the deformation of the Sigmoid function. The complete formula of the interval setting demand factor is: , is an independent variable. The present invention mainly considers the case where the independent variable is greater than 0. When the independent variable takes the value of 0, , minus 0.5 to filter out the influence of the independent variable on the overall calculation when it is 0 or less than 0, and illustratively, Usually, specific settings need to be made according to specific business scenarios, such as small transactions, daily consumption, etc. The value can be 500. For example, if the transaction amount fluctuates but is generally controllable, such as medium-sized transactions and regular transactions between enterprises, The value can be 5000.

[0048] It should be added that the number of intervals divided by the reference benchmark can be set based on experience or data rules, and is not specifically limited here. For example, the number of intervals divided by the reference benchmark is specifically set as follows: 1) Based on experience example

[0049] 1. Square root method: Take the cumulative number of transactions as the data point data, assuming the number of data points is , you can consider setting the number of partition intervals to Specifically, it can be set to ,Right now .

[0050] 2. Sturgis formula: The calculation formula for the number of divided intervals is: ,Right now , suppose there are 1000 data points, .

[0051] 2) Examples based on data patterns

[0052] 1. Information entropy method: Step 1: The transaction amount of each accumulated transaction constitutes a transaction set and calculates the information entropy of the data: , Indicates The transaction amount of the cumulative transaction, Indicates The probability of occurrence of the transaction amount of the cumulative transaction, represents information entropy, Indicates the cumulative transaction sequence number. .

[0053] Step 2: Divide the transaction amount interval into the number of randomly divided intervals, and calculate the information entropy after each division.

[0054] Step 3: Use the division order as the horizontal coordinate and the information entropy as the vertical coordinate to mark multiple points in the two-dimensional coordinate system, and construct a regression line for each marked point, and extract the slope of the regression line corresponding to each marked point. If the slope of the regression line corresponding to a certain marked point is greater than the set reference slope, and the slope of the regression line corresponding to the marked point before is less than the slope of the regression line corresponding to the marked point, then the division order to which the marked point belongs is used as the target division order, and the number of division intervals corresponding to the target division order is used as the reference benchmark division interval number.

[0055] It should be added that when the data is segmented, that is, after the interval is divided, when the information entropy is recalculated, it is necessary to determine the probability of data appearing in each new interval based on the newly divided intervals, and then substitute it into the formula for calculation. intervals, can represent the first interval, It is the probability that the transaction amount falls within the first interval. Similarly, by calculating the probability of each interval and substituting it into the information entropy formula, we can obtain the information entropy of the transaction set divided by intervals.

[0056] It should also be added that the example statistical process of the probability that the transaction amount falls within the first interval is: the number of transaction amount data points falling within the first interval is set to , It should be noted that this is not aimed at a specific transaction amount, but considers the distribution of all transaction amount data in different intervals, so as to calculate the probability of falling into a specific interval.

[0057] 2. Cluster Analysis

[0058] Use clustering algorithms, such as K-Means clustering, to perform cluster analysis on the transaction set, set different numbers of clusters, record the clustering results, and determine the appropriate number of clusters by evaluating the compactness and separation of the clusters.

[0059] It should be added that commonly used evaluation indicators include silhouette coefficient, Calinski-Harabasz index, etc. The value of silhouette coefficient is between -1 and 1, and the closer it is to 1, the better the clustering effect. The larger the Calinski-Harabasz index, the better the clustering effect. For example, try different numbers of clusters, calculate the corresponding evaluation indicators, and select the number of clusters with the largest index value as the reference benchmark for dividing the interval number. Cluster analysis is a relatively mature analysis technology, and its specific analysis process is not described in detail here.

[0060] Furthermore, an example of a specific division method for dividing the transaction amount interval into transaction amount sub-intervals in step A4 is as follows: Assume that we have the following transaction amount data: the highest transaction amount: 100,000 yuan, the lowest transaction amount: 1,000 yuan.

[0061] When the transaction amount interval division method is equal width interval division, and the number of divided intervals is 5, calculate the interval width: , that is, the transaction amount sub-intervals are: [1000, 20800), [20800, 40600), [40600, 60400), [60400, 80200), [80200, 100000].

[0062] When the transaction amount interval is divided into unequal width intervals, it is divided according to the data distribution: low amount interval: [1000, 5000), medium-low amount interval: [5000, 20000), medium amount interval: [20000, 50000), medium-high amount interval: [50000, 80000), high amount interval: [80000, 100000].

[0063] It should also be added that according to the data distribution, the transaction amount from each cumulative transaction can be imported into the number axis for distribution observation, or the market distribution rule can be defaulted. For the sake of ease of understanding and analysis, only an example is given here, and corresponding adjustments can be made in the actual process.

[0064] Further, in step A4, determining the target customer's preferred transaction amount range includes: taking the ratio of the cumulative number of transactions in each transaction amount sub-range to the recorded cumulative number of transactions as the cumulative transaction ratio of each transaction amount sub-range, recorded as , Indicates transaction amount sub-intervals, .

[0065] The accumulated transactions whose transaction dates are within the preset time window are recorded as adjacent transactions. The number of adjacent transactions within each transaction amount sub-interval is counted and compared with the accumulated number of transactions within the corresponding transaction amount sub-interval. The ratio is taken as the adjacent transaction ratio and recorded as .

[0066] Statistical transaction amount sub-interval transaction tendency , , and are the weights corresponding to the cumulative transaction ratio and the adjacent transaction ratio, , and take the transaction amount sub-range with the largest transaction propensity as the target customer's propensity transaction amount range.

[0067] It should be added that relying on any one indicator alone may overlook some important behavioral characteristics. For example, only looking at the cumulative transaction ratio may ignore those intervals where the transaction frequency is not high but there is continuous trading behavior, while only looking at the adjacent transaction ratio may ignore those intervals where there are no frequent and continuous transactions but there are many transactions in the long term. At the same time, the cumulative transaction ratio reflects the long-term preferences of customers' trading behavior, while the adjacent transaction ratio focuses more on short-term behavior patterns. By combining the two, we can better understand the behavioral habits of customers in different transaction amount ranges, understand the customer's behavior patterns more comprehensively, fully consider the factor of time, and avoid the shortcomings of selecting the preferred transaction amount range only from the frequency level.

[0068] It should be added that the cumulative transaction ratio refers to the ratio of the number of transactions occurring within a certain transaction amount sub-interval to the total number of transactions. The adjacent transaction ratio refers to the ratio of transactions occurring within a preset time window to the total number of transactions within the interval. Generally, companies pay more attention to the long-term transaction trends of their customers, so they can better avoid transaction risks. Therefore, the weight of the cumulative transaction ratio is set higher than that of the adjacent transaction ratio. For example, The value can be 0.6. The value can be 0.4.

[0069] Furthermore, regarding the analysis of the transaction fixity of the target customer in step A5, the following steps are included: A51. Compare the transaction dates of each cumulative transaction to obtain the transaction interval days of each cumulative transaction, and calculate the average transaction interval days by average calculation, which is recorded as .

[0070] A52. Calculate the standard deviation of the number of days between each cumulative transaction and record the result as , and then calculate the target customer's transaction time fixity , .

[0071] A53. Integrate the cumulative transactions in the same payment area, count the cumulative number of transactions in each payment area, and divide it by the recorded cumulative number of transactions to obtain the transaction ratio of each payment area. Calculate the payment area fixity of the target customer by combining the Gini coefficient. .

[0072] A54. Integrate the accumulated transactions of the same receiving account according to The calculation method is similar to the calculation method to obtain the fixed degree of the target customer's payment account. .

[0073] A55. Set the weights of transaction time, payment area and payment account, and obtain the transaction fixity of the target customer through weighted summation. .

[0074] It should be added that the fixed degree of trading time represents the periodicity fraction of trading dates. The closer the value is to 1, the stronger the transaction periodicity is, that is, the higher the fixed degree of transaction time is.

[0075] It should be added that the specific statistical formula for calculating the fixed degree of target customers' payment areas through the Gini coefficient is: ,in, The calculation formula for the Gini coefficient is: Indicates The transaction number ratio of each receiving region is Indicates the payment area code. Because the Gini coefficient value range is [0, 1], and the closer it is to 0, the greater the fixed degree of the collection area. In order to ensure the consistency of the subsequent overall statistical results and the denominator is not 0, a compensation of 1 is made in the denominator, and the inverse of the Gini coefficient is taken as the final result of the fixed degree of the collection area.

[0076] It should also be added that when calculating the fixity of the payment area and the fixity of the payment account, the Herfindahl index can also be used for calculation. For example, the reference formula for calculating the fixity of the payment area using the Herfindahl index is: , where the Herfindahl index ranges from 0 to 1. The more dispersed the payment areas are, the closer the Herfindahl index is to 0, and the more concentrated the payment areas are, the closer the Herfindahl index is to 1. The calculation principle of the payment area fixity and the payment account fixity are the same. The specific calculation formula of the payment account fixity can be derived by analogy, and no example is given here.

[0077] In a specific embodiment, the transaction time takes into account the stability and regularity of the transaction time, and thus the maximum weight is set. Specifically, the weight of the transaction time may be 0.4, and the payment area and payment account both reflect the attribute weights of the transaction object. Therefore, the payment area and payment account may be given the same weight, that is, the weights of the payment area and payment account are both 0.3.

[0078] The embodiment of the present invention obtains a comprehensive fixity index by comprehensively considering the fixity of three dimensions, namely, transaction time, transaction region and transaction account, and setting appropriate weights according to business needs. This helps to more comprehensively understand the customer's transaction behavior patterns, solves the current problem of insufficient control over changes in the customer's time dimension, and further improves the accuracy of subsequent assessments of whether or not early warnings are triggered.

[0079] See also Figure 3 As shown, the warning trigger assessment module is used to assess whether the warning is triggered based on the current transaction data to be confirmed and the transaction feature information. If not, the warning instruction is not started. If so, the warning trigger confirmation module is started.

[0080] The evaluation of whether the early warning is triggered includes: S1, extracting the transaction amount to be confirmed, the payment area and the payment account number of the transaction amount to be confirmed from the current transaction data to be confirmed.

[0081] S2. Determine whether the transaction amount to be confirmed exceeds the maximum transaction amount of the target customer. If so, it will be used as the evaluation result. Otherwise, execute step S3.

[0082] S3. Compare the transaction amount to be confirmed with the preferred transaction amount range. If the transaction amount to be confirmed is within the preferred transaction amount range or is less than the lower limit of the preferred transaction amount range, confirm the security of the collection and execute step S4. If the transaction amount to be confirmed is higher than the upper limit of the preferred transaction amount range, set the security exceeding the preferred transaction amount value, calculate the difference between the transaction amount to be confirmed and the upper limit of the preferred transaction amount range, and execute step S5.

[0083] S4. Determine whether the security level of the payment is greater than or equal to the set reference security level. If so, take “no” as the evaluation result; if not, take “yes” as the evaluation result.

[0084] It should be added that the reference safety degree is set to be a real number slightly greater than 0. For example, the specific value may be 0.1.

[0085] S5. Determine whether the difference is greater than the safe excess tendency transaction amount value. If so, it will be used as the assessment result. If not, return to step S4.

[0086] Furthermore, regarding the confirmation of payment security in step S3, it includes: J1. Recording the payment area and payment account of the transaction amount to be confirmed as the area to be confirmed and the account to be confirmed respectively.

[0087] J2. Mark the payment collection areas of each cumulative transaction on the electronic map to obtain the total number of markings, the number of markings of each marked payment collection area, and the transaction date of each marking. The ratio of the number of markings of each marked payment collection area to the total number of markings is recorded as the marking ratio.

[0088] J3. If the area to be confirmed is inconsistent with the marked payment areas, 0 is used as the payment security level. Otherwise, the corresponding marked transaction dates of the area to be confirmed are extracted.

[0089] J4. Count the number of times the transaction date is marked within the preset time window and compare it with the total number of times marked to obtain the effective marking ratio of the area to be confirmed, recorded as .

[0090] J5. The payment account number and transaction date of each accumulated transaction constitute a payment list. If the account to be confirmed is not in the payment list, 0 is used as the payment security level. Otherwise, the transaction date of each accumulated transaction corresponding to the account to be confirmed is extracted from the payment account list.

[0091] J6. The accumulated transactions whose transaction dates are within the preset time window are recorded as valid transactions. The number of valid transactions is counted and compared with the number of accumulated transaction dates recorded to obtain the valid marking ratio of the account to be confirmed, which is recorded as ,Will As the security of receiving payment, it is recorded as , and The effective marking ratio of the region and the effective marking ratio of the account are set as the reference respectively, so as to obtain the payment security, which is recorded as , The value is 0 or .

[0092] It should be added that and The specific value of is set as follows: The analysis method of the effective marking ratio of the area to be confirmed is similar to the analysis to obtain the effective marking ratio of each marked collection area, calculate the average of the effective marking ratio of each marked collection area, and use the calculation result as The specific value of .

[0093] The receiving accounts of each cumulative transaction are deduplicated to obtain the remaining receiving accounts after deduplication. The valid marking ratios of the remaining receiving accounts after deduplication are analyzed in the same way as the valid marking ratios of the accounts to be confirmed, and the average is calculated. The calculation result is used as The specific value of .

[0094] It should also be added that the payment security is set with reference to the ReLU function. The final selection of the minimum value is to ensure the reference and effectiveness of the payment security, to avoid the occurrence of a certain degree of offsetting effect between the effective regional marking ratio and the effective account marking ratio when performing mean calculation or weight distribution, and to ensure the representativeness of the payment security evaluation results.

[0095] Furthermore, setting the safe excess tendency transaction amount value in step S3 includes: B1, extracting the transaction tendency type from the transaction feature information of the target user.

[0096] B2. If the trading trend type is growth, the excess factor is recorded as , if it is a fluctuation, the excess factor is recorded as , if it is robust, the excess factor is recorded as , , which sets the excess factor , The value is or or .

[0097] B3. Record the upper limit of the preferred transaction amount range and the maximum transaction amount as and , set the safe excess tendency transaction amount value, recorded as , , and Represent the proportion of transaction fixity and excess factor respectively.

[0098] It should be added that when the transaction trend type is growth, it means that the customer is constantly increasing the transaction amount, and the acceptable excess range is higher. Therefore, the excess factor value is also the largest. When the transaction trend type is fluctuation, there is a process of floating up and down on the surface, but the increase is not large and the regularity is not clear. Setting the excess factor is second. When the transaction trend type is stable, it cannot directly reflect the acceptable excess range. Therefore, the excess factor is set to the minimum. For example, , , The values ​​may be 0.5, 0.3 and 0.2 respectively, and in a specific embodiment, The specific value can be 0.55. The specific value can be 0.45.

[0099] When evaluating whether an early warning is triggered, the embodiment of the present invention makes a comprehensive consideration of the three dimensions of amount, payment area, and payment account, and sets a transaction amount value with a safety excess tendency and a confirmation of payment safety. This allows a more comprehensive assessment of transaction risks and reduces security vulnerabilities that may be missed in a single dimension. It also helps to reduce false alarms and missed alarms, improve the overall reliability, sensitivity, and accuracy of early warnings, and thereby improve customer transaction security.

[0100] The warning trigger confirmation module is used to confirm the warning risk category of the target customer.

[0101] Specifically, the warning risk category of the confirmed target customer includes: if the transaction amount to be confirmed exceeds the target customer's maximum transaction amount, or the difference between the transaction amount to be confirmed and the corresponding upper limit value of the tendency transaction amount range exceeds the set safety excess tendency transaction amount value, the amount risk is taken as a warning risk category.

[0102] If the transaction amount to be confirmed is within the preferred transaction amount range or is less than the lower limit of the preferred transaction amount range, and the collection security is less than the set reference security, the collection risk will be regarded as a warning risk category.

[0103] If the transaction amount to be confirmed is higher than the upper limit of the preferred transaction amount range, and the difference with the corresponding upper limit of the preferred transaction amount range exceeds the set safety excess preferred transaction amount value, and at the same time the collection security is less than the set reference security, the amount risk and collection risk will be taken as warning risk categories.

[0104] The embodiment of the present invention evaluates whether the warning is triggered by combining the customer's transaction feature information and confirms the warning risk category, which is convenient for subsequent precise and targeted warnings, avoids the shortcomings of the current general and more common warning methods, fully considers the customer's transaction characteristics and other laws, realizes the customer's detailed warning evaluation, and ensures the timeliness and accuracy of subsequent transaction warnings.

[0105] The risk warning feedback terminal is used to send corresponding warning instructions to the target customer's transaction Web interface based on the warning risk category of the target customer.

[0106] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they shall all fall within the protection scope of the present invention.

Claims

1. A customer potential risk intelligent early warning system based on SAAS technology, characterized in that: The system includes: The transaction data collection module is used to record the current transaction customer as the target customer and collect the target customer's tracking transaction data and the current transaction data to be confirmed from the cloud; The transaction data analysis module is used to analyze the tracking transaction data of target customers and obtain the transaction feature information of target users; The warning trigger assessment module is used to assess whether the warning is triggered based on the current transaction data to be confirmed and the transaction feature information. If not, the warning instruction is not activated. If so, the warning trigger confirmation module is activated; The early warning trigger confirmation module is used to confirm the early warning risk category of the target customer; The risk warning feedback terminal is used to send corresponding warning instructions to the target customer's transaction web interface based on the target customer's warning risk category; The analysis of the tracking transaction data of the target customers includes: Extract the cumulative number of transactions and the transaction amount, transaction date, payment region and payment account number of each cumulative transaction from the tracking transaction data; Identify the target customers' trading trend types, including growth, volatility and stability; Filter out the highest transaction amount and the lowest transaction amount from the accumulated transaction amounts of each transaction to construct a transaction amount range; Select the transaction amount interval division method and the number of intervals, divide the transaction amount interval into each transaction amount sub-interval, count the cumulative number of transactions within each transaction amount sub-interval, and confirm the target customer's preferred transaction amount interval; Based on the payment area and payment account, the transaction fixity of the target customer is analyzed and recorded as ; The transaction trend type, maximum transaction amount, preferred transaction amount range and transaction fixity are used as the transaction characteristic information of the target user; The analysis of the target customer's transaction fixity includes: Compare the transaction dates of each cumulative transaction to obtain the transaction interval days of each cumulative transaction. The average transaction interval days is obtained by average calculation, which is recorded as ; Calculate the standard deviation of the number of days between each cumulative transaction and record the result as , and then calculate the target customer's transaction time fixity , ; Integrate the cumulative transactions in the same payment area, count the cumulative number of transactions in each payment area, and divide it by the recorded cumulative number of transactions to get the transaction ratio of each payment area. Calculate the fixed degree of payment area of ​​the target customer by combining the Gini coefficient. ; Integrate the accumulated transactions of the same payment account into The calculation method is similar to the calculation method to obtain the fixed degree of the target customer's payment account. ; Set the weights of transaction time, payment area and payment account, and obtain the transaction fixity of the target customer through weighted summation. ; The assessment of whether the early warning is triggered includes: S1. Extract the transaction amount to be confirmed, the payment area and payment account number of the transaction amount to be confirmed from the current transaction data to be confirmed; S2, determine whether the transaction amount to be confirmed exceeds the maximum transaction amount of the target customer, if so, it will be used as the assessment result, otherwise, execute step S3; S3, compare the transaction amount to be confirmed with the preferred transaction amount interval. If the transaction amount to be confirmed is within the preferred transaction amount interval or is less than the lower limit of the preferred transaction amount interval, confirm the safety of the payment and execute step S4. If the transaction amount to be confirmed is higher than the upper limit of the preferred transaction amount interval, set the safety exceeding the preferred transaction amount value, calculate the difference between the transaction amount to be confirmed and the upper limit of the preferred transaction amount interval, and execute step S5. S4, judging whether the security level of the payment is greater than or equal to the set reference security level, if so, taking “no” as the evaluation result, if not, taking “yes” as the evaluation result; S5, judging whether the difference is greater than the safe excess tendency transaction amount value, if so, it will be used as the assessment result, if not, returning to step S4; The confirmation of payment security includes: The receiving area and receiving account number of the transaction amount to be confirmed are recorded as the area to be confirmed and the account number to be confirmed respectively; Mark the payment collection areas of each accumulated transaction on the electronic map to obtain the total number of markings, the number of markings of each marked payment collection area, and the transaction date of each marking, and record the ratio of the number of markings of each marked payment collection area to the total number of markings as the marking ratio; If the area to be confirmed is inconsistent with the marked payment areas, 0 is used as the payment security level. Otherwise, the corresponding marked transaction dates of the area to be confirmed are extracted. Count the number of times the transaction date is marked within the preset time window and compare it with the total number of marks to obtain the effective marking ratio of the area to be confirmed, which is recorded as ; The receiving account and transaction date of each accumulated transaction form a receiving list. If the account to be confirmed is not in the receiving list, 0 is used as the receiving security level. Otherwise, the transaction date of each accumulated transaction corresponding to the account to be confirmed is extracted from the receiving account list. The accumulated transactions whose transaction dates are within the preset time window are recorded as valid transactions. The number of valid transactions is counted and compared with the number of accumulated transaction dates recorded to obtain the valid marking ratio of the account to be confirmed, which is recorded as ,Will As the security of receiving payment, it is recorded as , and The effective marking ratio of the region and the effective marking ratio of the account are set as the reference respectively, so as to obtain the payment security, which is recorded as , The value is 0 or .

2. The SAAS-based intelligent early warning system for potential customer risks as claimed in claim 1, characterized in that: The transaction trend types of the confirmed target customers include: With the accumulated transaction order as the horizontal axis and the transaction amount as the vertical axis, a transaction amount curve is constructed, and the slope is extracted from it as the transaction growth rate. At the same time, the number of fluctuation points is extracted and compared with the accumulated number of transactions recorded, and the ratio is used as the fluctuation ratio; If the transaction growth rate is greater than or equal to the set reference transaction growth rate as condition 1, the volatility ratio is less than or equal to the set volatility ratio as condition 2; If condition 1 and condition 2 are met, growth will be used as the trading trend type. If only condition 2 is met, stability will be used as the trading trend type. If only condition 1 is met, volatility will be used as the trading trend type.

3. The SAAS-based intelligent early warning system for potential customer risks as claimed in claim 1, characterized in that: The target customer's preferred transaction amount range includes: The ratio of the cumulative number of transactions in each transaction amount sub-interval to the cumulative number of transactions recorded is taken as the cumulative transaction ratio of each transaction amount sub-interval, denoted as , Indicates transaction amount sub-intervals, ; The accumulated transactions whose transaction dates are within the preset time window are recorded as adjacent transactions. The number of adjacent transactions within each transaction amount sub-interval is counted and compared with the accumulated number of transactions within the corresponding transaction amount sub-interval. The ratio is taken as the adjacent transaction ratio and recorded as ; Statistical transaction amount sub-interval transaction tendency , , and are the weights corresponding to the cumulative transaction ratio and the adjacent transaction ratio, , and take the transaction amount sub-range with the largest transaction propensity as the target customer's propensity transaction amount range.

4. The SAAS-based intelligent early warning system for potential customer risks as claimed in claim 1, characterized in that: The settings mentioned above safely exceed the preferred transaction amount value, including: Set the excess factor, denoted as ; The upper limit of the preferred transaction amount range and the maximum transaction amount are recorded as and , set the safe excess tendency transaction amount value, recorded as , , and Represent the proportion of transaction fixity and excess factor respectively.

5. The SAAS-based intelligent early warning system for potential customer risks as claimed in claim 4, characterized in that: The settings exceed the factors, including: Extracting transaction trend types from transaction feature information of target users; If the trading trend type is growth, the excess factor is recorded as , if it is a fluctuation, the excess factor is recorded as , if it is robust, the excess factor is recorded as , , which sets the excess factor , The value is or or .

6. The SAAS-based intelligent early warning system for potential customer risks as claimed in claim 1, characterized in that: The warning risk categories for confirming target customers include: If the transaction amount to be confirmed exceeds the maximum transaction amount of the target customer, or the difference between the transaction amount to be confirmed and the corresponding upper limit of the preferred transaction amount range exceeds the set safe excess preferred transaction amount value, the amount risk will be regarded as a warning risk category; If the transaction amount to be confirmed is within the range of the preferred transaction amount or is less than the lower limit of the range of the preferred transaction amount, and the collection security is less than the set reference security, the collection risk is regarded as a warning risk category; If the transaction amount to be confirmed is higher than the upper limit of the preferred transaction amount range, and the difference with the corresponding upper limit of the preferred transaction amount range exceeds the set safety excess preferred transaction amount value, and at the same time the collection security is less than the set reference security, the amount risk and collection risk will be taken as warning risk categories.

Images