Method, device and electronic equipment for monitoring the use of funds of a consumer loan

Abstract

The application provides a fund use monitoring method and device for consumer loans and electronic equipment. In the scheme, a first clustering algorithm is used to extract a plurality of typical daily use curves of each user in different time periods from the obtained user credit use data, and a typical daily use curve of a user group is determined according to the typical daily use curves; the real-time daily use curve of each user on the use application day is compared with the plurality of typical daily use curves of the user in similarity, and the minimum similarity comparison result is selected; the real-time daily use curve of each user is compared with the typical daily use curve of the user group in trend, and the maximum trend comparison result is selected; the minimum similarity comparison result and the maximum trend comparison result are compared with the corresponding preset threshold value respectively, and whether the use behavior of the user is abnormal is monitored. The scheme solves the problem that the identification ability of abnormal fund use risk is limited in the use application stage of consumer loans in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis and processing, in particular to a fund use monitoring method and device for consumer loans, a computer readable storage medium and an electronic device. BACKGROUND

[0002] In the prior art, a graph theory model or a decision tree model is used to monitor the fund flow direction of a loan. This method only realizes fund flow direction monitoring after the loan, and cannot expose risks in advance, that is, cannot completely recover abnormal transactions that have occurred, which will cause unavoidable economic losses. In addition, in the prior art, risk assessment of loan use is based on historical data of accounts and account groups. The historical data used has a short value period and is accidental, and cannot accurately represent the behavior habits of target accounts and account groups. In addition, the determination of the account group in the method is only based on the interval of the use amount of the account applied for this time, and the judgment factor is single, so it cannot accurately identify abnormal fund use risks. SUMMARY

[0003] The main purpose of the present application is to provide a fund use monitoring method, device, computer readable storage medium and electronic device for consumer loans, to at least solve the problem of limited identification ability of abnormal fund use risks in the application stage of consumer loan use in the prior art.

[0004] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a fund use monitoring method for consumer loans is provided, comprising: acquiring user quota use data of a plurality of users, and using a first clustering algorithm to perform clustering analysis on the plurality of user quota use data, and extracting a plurality of typical daily use curves of each user in different time periods, the time periods including a time period composed of a plurality of working days in a preset time period, and a time period composed of a plurality of holidays in the preset time period; determining a typical daily use curve of a user group according to the typical daily use curves of all users; comparing a real-time daily use curve of each user on the day of use application with the plurality of typical daily use curves of the user to obtain a plurality of similarity comparison results, and selecting a minimum similarity comparison result, the minimum similarity comparison result being the minimum value in the plurality of similarity comparison results; comparing the real-time daily use curve of each user with the typical daily use curve of the user group to obtain a plurality of trend comparison results, and selecting a maximum trend comparison result, the maximum trend comparison result being the maximum value in the plurality of trend comparison results; comparing the minimum similarity comparison result and the maximum trend comparison result with corresponding preset thresholds respectively, and monitoring whether the use behavior of the user is abnormal.

[0005] According to another aspect of the present application, a fund use monitoring device for consumer loans is provided, comprising: an acquisition unit configured to acquire user credit usage data of a plurality of users and perform cluster analysis on the plurality of user credit usage data using a first clustering algorithm to extract a plurality of typical daily usage curves of each of the users in different time periods, the time periods including a time period formed by a plurality of workdays within a preset time period and a time period formed by a plurality of holidays within the preset time period; a first determination unit configured to determine a typical daily usage curve of a user group according to the typical daily usage curves of all the users; a similarity comparison unit configured to compare a real-time daily usage curve of each of the users on a usage day with the plurality of typical daily usage curves of the user to obtain a plurality of similarity comparison results and select a minimum similarity comparison result, the minimum similarity comparison result being the minimum value among the plurality of similarity comparison results; a trend comparison unit configured to compare the real-time daily usage curve of each of the users with the typical daily usage curve of the user group to obtain a plurality of trend comparison results and select a maximum trend comparison result, the maximum trend comparison result being the maximum value among the plurality of trend comparison results; and a comparison unit configured to compare the minimum similarity comparison result and the maximum trend comparison result with corresponding preset thresholds, respectively, to monitor whether the usage behavior of the user is abnormal.

[0006] According to still another aspect of the present application, a computer readable storage medium is provided, the computer readable storage medium including a stored program, wherein the program, when executed, controls a device in which the computer readable storage medium is located to perform any of the fund use monitoring methods for consumer loans.

[0007] According to yet another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include any of the fund use monitoring methods for consumer loans.

[0008] The technical scheme of the application is applied to obtain user credit consumption data of a plurality of users, and a first clustering algorithm is used to perform clustering analysis on the plurality of user credit consumption data, and a plurality of typical daily consumption curves of each user in different time periods are extracted, the time periods including a time period formed by a plurality of workdays in a preset time period and a time period formed by a plurality of holidays in the preset time period; a typical daily consumption curve of a user group is determined according to the typical daily consumption curves of all users; a real-time daily consumption curve of each user on the application consumption day is compared with the plurality of typical daily consumption curves of the user in terms of similarity, a plurality of similarity comparison results are obtained, and a minimum similarity comparison result is selected, the minimum similarity comparison result being the minimum value in the plurality of similarity comparison results; the real-time daily consumption curve of each user is compared with the typical daily consumption curve of the user group in terms of trend, a plurality of trend comparison results are obtained, and a maximum trend comparison result is selected, the maximum trend comparison result being the maximum value in the plurality of trend comparison results; the minimum similarity comparison result and the maximum trend comparison result are compared with corresponding preset thresholds respectively, and whether the consumption behavior of the user is abnormal is monitored. The scheme immediately performs risk assessment on the real-time consumption application data of the user on the application consumption day, overcomes the hysteresis of the traditional post-loan fund monitoring in the prior art, extracts the consumption behavior habit curve of the user and the user group based on a large amount of historical user credit consumption data, makes the extracted comparison curve more referential, can more accurately predict whether the current consumption of the user is abnormal, and thus solves the problem of limited identification ability of abnormal fund use risk in the consumption loan consumption application stage in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0009] The accompanying drawings, which form a part of the specification, are included to provide a further understanding of the application and are incorporated herein in conjunction with the description of the application. The embodiments of the present application, and their

[0010] Figure 1 A hardware structure block diagram of a mobile terminal for performing a fund use monitoring method of a consumption loan is shown according to an embodiment of the application;

[0011] Figure 2 A flowchart of a fund use monitoring method of a consumption loan is shown according to an embodiment of the application;

[0012] Figure 3 A flowchart of a specific fund use monitoring method of a consumption loan is shown according to an embodiment of the application;

[0013] Figure 4 A structure block diagram of a fund use monitoring device of a consumption loan is shown according to an embodiment of the application.

[0014] Wherein, the above-mentioned drawings include the following reference signs:

[0015] 102, processor; 104, memory; 106, transmission device; 108, input and output device. DETAILED DESCRIPTION

[0016] It should be noted that the embodiments and features in the present application can be combined with each other without conflict. The present application will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0017] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0018] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, not necessarily to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0019] As introduced in the background, the prior art usually only monitors the flow of funds after the loan, and the historical data used has a short value period and a single judgment factor, which cannot accurately identify abnormal fund use risks. To solve the problem that the prior art has limited ability to identify abnormal fund use risks in the application stage of consumer loan use, the embodiments of the present application provide a fund use monitoring method, device, computer readable storage medium and electronic equipment for consumer loans.

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0021] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking the case of running on a mobile terminal, Figure 1 is a hardware structure block diagram of a mobile terminal of a fund use monitoring method for consumer loans according to an embodiment of the present application. As shown inFigure 1 As shown, a mobile terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. The mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0022] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the device information display method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof. The transmission device 106 is used to receive or send data via a network. Specific examples of the aforementioned networks may include wireless networks provided by the mobile terminal's communication provider. In one example, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0023] This embodiment provides a method for monitoring the use of consumer loan funds that runs on a mobile terminal, computer terminal, or similar computing device. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0024] Figure 2is a flowchart of a method for monitoring the use of funds of a consumer loan according to an embodiment of the present application. As shown in Figure 2 the method comprises the following steps:

[0025] In step S201, user credit usage data of a plurality of users is obtained, and a first clustering algorithm is used to perform clustering analysis on the plurality of user credit usage data, and a plurality of typical daily usage curves of each user in different time periods are extracted, wherein the time periods include a time period formed by a plurality of workdays within a preset time period, and a time period formed by a plurality of holidays within the preset time period.

[0026] Specifically, user credit usage data of a plurality of users is obtained from a related database. These user credit usage data contain the consumption loan usage records of each user at each time point (such as every day), including the amount of usage, time stamp and other information. The time range covered by the data can be one year, two years or a longer period of time to ensure that there is enough historical data for analysis. A first clustering algorithm is used to perform clustering analysis on the plurality of user credit usage data. In this embodiment, the first clustering algorithm refers to the K-means++ clustering algorithm, and the purpose is to group the usage behavior of users and analyze the usage habits of each user in different time periods. Clustering analysis is based on the characteristics of user usage data, such as usage time, usage amount, etc., and data points are assigned to different clusters (or groups), each cluster representing a usage behavior pattern. Through clustering, the consumption habits and patterns of each user in different time periods can be revealed, providing a basis for subsequent risk assessment.

[0027] Through cluster analysis, a curve representing the average spending behavior of users within each cluster is extracted, known as the typical daily spending curve. This curve reflects the average spending pattern of users across different time periods (weekdays and holidays). Specifically, since holiday spending patterns differ significantly from weekday spending patterns, cluster analysis of users' historical spending data identifies clusters with similar spending habits from a large pool of historical data. The cluster center of each cluster is then calculated, and this cluster center forms the typical daily spending curve. Users' spending habits differ between weekdays and holidays. Weekdays can be further subdivided into the beginning, middle, and end of the month, while holidays can be subdivided into weekends, long holidays, short holidays, and shopping festivals. Users' spending data from the past year can be obtained, encompassing various weekday and holiday scenarios. By performing cluster analysis on these user credit limit usage data, typical daily spending curves that represent the user's spending behavior characteristics across different time periods are automatically identified and extracted. These typical daily spending curves are automatically generated based on naturally formed clustering patterns in the user's historical spending data. For example, the cluster analysis results may reveal a curve that particularly matches the user's spending behavior on most weekdays; another curve represents the user's spending behavior on weekends or public holidays; and yet another curve reflects the user's spending behavior during shopping festivals. By using cluster analysis to identify users' typical daily spending curves, it is possible to more accurately capture the user's spending behavior characteristics across different time periods, thereby better identifying potential abnormal spending behavior.

[0028] Step S202: Based on the typical daily spending curves of all the above-mentioned users, determine the typical daily spending curve of the user group.

[0029] Specifically, the purpose of step S202 is to further extract typical daily spending curves for user groups from the typical daily spending curves of all users. In step S201, cluster analysis was performed on the user credit spending data of each user, thereby extracting multiple typical daily spending curves for each user in different time periods. These typical daily spending curves reflect the consumption patterns and habits of individual users based on their historical data. Furthermore, in order to more comprehensively assess whether there are any anomalies in user spending behavior, it is not enough to consider only the typical daily spending curves of individual users; it is also necessary to consider the behavior of a broader user group. Therefore, it is necessary to extract typical daily spending curves for user groups based on the typical daily spending curves of all users. By dividing users into different groups and further analyzing the typical daily spending curves of all users within each group, curves that can represent the consumption patterns of that group are obtained, i.e., the typical daily spending curves of the user group. The typical daily spending curves of each group reflect the common consumption habits of users in that group. In summary, the purpose of extracting typical daily spending curves for user groups is to establish a benchmark, provide a normal consumption pattern for the group to which an individual user belongs, and provide a reference for judging abnormal spending behavior of users in the future.

[0030] Step S203: Compare the real-time daily spending curve of each user on the day of application with the multiple typical daily spending curves of the user to obtain multiple similarity comparison results, and select the minimum similarity comparison result. The minimum similarity comparison result is the minimum value among the multiple similarity comparison results.

[0031] Specifically, the user's real-time daily spending curve is compared with each of the user's typical daily spending curves to obtain multiple similarity comparison results, quantifying the degree of similarity between the user's real-time daily spending curve and each of the user's typical daily spending curves. The result with the lowest similarity is selected from these multiple similarity comparison results. The result with the lowest similarity means that the difference between the user's real-time daily spending curve and a particular typical daily spending curve is the smallest, that is, the user's real-time spending behavior is closest to the user's typical spending behavior pattern. By selecting the result with the lowest similarity, a quantitative indicator can be provided for subsequent judgment of abnormal user behavior.

[0032] Step S204: Compare the real-time daily spending curves of each user with the typical daily spending curves of the user group to obtain multiple trend comparison results, and select the maximum trend comparison result. The maximum trend comparison result is the maximum value among the multiple trend comparison results.

[0033] Specifically, the real-time daily spending curve on the day a user applies for a loan refers to the change in the actual amount spent by the user over time on the day they apply for the consumer loan. By comparing each user's real-time daily spending curve with the typical daily spending curves of multiple user groups identified beforehand, multiple trend comparison results can be obtained. These trend comparison results reflect the difference in trend between the user's consumption behavior on the day of the loan application and the consumption behavior of their group. The result with the strongest trend comparison is selected from these multiple results. This result represents the highest correlation, meaning the user's real-time daily spending curve most closely matches the trend of a typical daily spending curve of a particular user group, indicating the most consistent real-time behavior between the user and the historical behavior of their group. By selecting the result with the strongest trend comparison, a quantitative indicator can be provided for subsequent judgment of abnormal user behavior.

[0034] Step S205: Compare the minimum similarity comparison result and the maximum trend comparison result with the corresponding preset thresholds to monitor whether there are any abnormalities in the user's spending behavior.

[0035] Specifically, the maximum trend comparison result is compared with the corresponding preset threshold, which is obtained through simulation training with a large amount of actual data. This preset threshold is used to determine whether the trend of user behavior is normal. If the maximum trend comparison result is lower than the preset threshold, it indicates that the user's current spending trend is significantly different from the group behavior, indicating that the user's spending behavior is abnormal. Similarly, the minimum similarity comparison result is compared with the corresponding preset threshold, which is also obtained through simulation training with a large amount of actual data. If the minimum similarity comparison result exceeds the preset threshold, it indicates that the user's current spending behavior deviates significantly from the typical spending pattern in the past, which also indicates that there is an abnormal spending behavior.

[0036] By comparing the minimum similarity comparison results and the maximum trend comparison results with corresponding preset thresholds, it is possible to determine whether there are any anomalies in user spending behavior. The selection of these preset thresholds is crucial; they are set based on extensive historical data and a deep understanding of the user group, aiming to ensure that abnormal spending behavior is captured without misjudging normal spending activities. Using the minimum similarity comparison results and the maximum trend comparison results in conjunction with the preset thresholds, anomalies in user spending behavior can be monitored in real time, potential risks can be identified promptly, and the risk identification capability for fund usage can be improved.

[0037] This embodiment enables immediate risk assessment of users' real-time withdrawal application data on the day of application, overcoming the lag in traditional post-loan fund monitoring in existing technologies. Furthermore, it extracts withdrawal behavior curves for users and user groups based on a large amount of historical user credit limit withdrawal data, making the extracted comparison curves more reliable and allowing for more accurate prediction of whether users' current withdrawals are abnormal. This solves the problem of limited ability to identify abnormal fund usage risks in the consumer loan withdrawal application stage in existing technologies.

[0038] In the specific implementation process, based on the typical daily spending curves of all the aforementioned users, the typical daily spending curves of the user group are determined, including: using a second clustering algorithm to perform cluster analysis on the typical daily spending curves of all the aforementioned users to obtain multiple clusters of typical daily spending curves; and determining the cluster center curves of the multiple clusters of typical daily spending curves as the typical daily spending curves of the aforementioned user group. Specifically, using the second clustering algorithm to perform cluster analysis on the typical daily spending curves of all the aforementioned users to obtain multiple clusters of typical daily spending curves includes: using the second clustering algorithm, for each of the aforementioned users, taking the typical daily spending curves of each user as the initial centroid, calculating the Euclidean distances between the typical daily spending curves of all other users and the initial centroids, obtaining multiple Euclidean distances for each of the aforementioned users; assigning the typical daily spending curves of each of the aforementioned users to the cluster containing the initial centroid corresponding to the smallest Euclidean distance among the multiple Euclidean distances for that user, recalculating the centroid of each of the aforementioned clusters until a preset stopping condition is reached, thus obtaining multiple clusters of typical daily spending curves for each of the aforementioned users.

[0039] The above content further details how to determine the typical daily spending curves of a user group. Specifically, after obtaining the typical daily spending curves of all users, a second clustering algorithm is used to cluster all typical daily spending curves. In this embodiment, the second clustering algorithm is K-means++ clustering. The purpose of clustering is to group curves with similar consumption behavior patterns together to form multiple typical daily spending curve clusters. The clustering algorithm automatically groups curves based on the similarity or difference between them. Curves within each group have high similarity in spending behavior, while curves between different groups show significant differences in behavior patterns. The typical daily spending curve of each user is used as the initial centroid, meaning that each user's typical daily spending curve is considered a cluster center, which helps ensure that the algorithm fully considers the historical behavior patterns of each user. For example, when determining the user group to which a user belongs, the typical daily spending curve of that user is used as the initial centroid. Then, the Euclidean distance between each typical daily spending curve of other users and the initial centroid of that user is calculated, resulting in multiple Euclidean distances for each user. Euclidean distance is a method for measuring the spatial distance between two points. In this embodiment, it is used to quantify the differences between curves; the smaller the distance, the more similar the two curves are. Then, each user's typical daily spending curve is assigned to the cluster containing the initial centroid corresponding to it with the smallest Euclidean distance. In this way, similar curves cluster together, forming preliminary clusters. This process essentially automatically categorizes the historical spending behavior of a user group, with each category representing the consumption habits of that user group during a specific period.

[0040] Whenever a user's typical daily spending curve is analyzed and added, it is assigned to the closest cluster based on its similarity to existing curves, and the centroid of the assigned cluster is recalculated. The new centroid can be the average of all curves within the cluster. The entire process (calculating Euclidean distance, assigning curves, and updating centroids) is then repeated until a preset stopping condition is met. In this embodiment, the preset stopping condition can be that the centroid no longer changes, the change in the centroid is less than a certain threshold, or a preset number of iterations has been reached, indicating that the cluster division has stabilized and no further adjustment is needed.

[0041] Through the above process, multiple typical daily spending curve clusters are ultimately obtained, each cluster containing a set of curves with similar spending behaviors. The cluster center curve reflects the typical spending behavior of the user group to which the user belongs within a certain time period. Determining the cluster center curve allows for clear identification of the spending behavior patterns of the user group to which the user belongs in different time periods, providing a benchmark for subsequent comparison of group behavior trends. Finally, the cluster center curve of each user's typical daily spending curve cluster is determined as the typical daily spending curve of that user group. Each typical daily spending curve of a user group represents the typical consumption behavior of a type of user within a specific time period. These typical daily spending curves of user groups will be used to compare trends with the real-time daily spending curve of an individual user on the day they apply for spending, thereby assessing whether the user's behavior deviates from the normal pattern of their group and identifying potential abnormal spending behavior.

[0042] In some embodiments, before performing cluster analysis on the typical daily spending curves of all the aforementioned users using the second clustering algorithm, the method includes: determining the initial number of cluster centers using the elbow rule, and comparing the initial number of cluster centers with the number of typical daily spending curves of all the aforementioned users; if the initial number of cluster centers is greater than or equal to the number of typical daily spending curves of all the aforementioned users, determining the initial number of cluster centers as the number of cluster centers; if the initial number of cluster centers is less than the number of typical daily spending curves of all the aforementioned users, determining the number of typical daily spending curves of all the aforementioned users as the number of cluster centers.

[0043] The above describes the process of determining the number of cluster centers before analyzing the clustering algorithm. First, the elbow rule is used to determine a reasonable initial number of cluster centers. The elbow rule is a method for finding the optimal k-value for data clustering. It selects an inflection point as the k-value by calculating the clustering cost function (such as the sum of squared errors within clusters) under different k-values. This inflection point is usually characterized by a significant slowdown in the rate of decrease of the cost function, meaning that adding more cluster centers reduces the marginal benefit of further reducing errors. After obtaining the initial number of cluster centers through the elbow rule, it is compared with the number of typical daily spending curves for all users. The number of typical daily spending curves reflects the number of users' historical behavior data clustered into typical curves. When the initial number of cluster centers determined by the elbow rule is greater than or equal to the number of typical daily spending curves for all users, this initial number of cluster centers is used as the total number of cluster centers. This is because a sufficient number of cluster centers ensures that each typical daily spending curve belongs to at least one cluster, avoiding missing clusters. If the initial number of cluster centers is less than the number of typical daily spending curves for all users, the number of typical daily spending curves for all users will be used as the number of cluster centers. This step ensures that the typical daily spending curve of each user can be taken into account, and avoids the omission of certain user behavior patterns due to insufficient number of cluster centers.

[0044] By following the above process, before starting cluster analysis, based on the elbow rule and the number of typical daily spending curves for all users, a number of cluster centers can be selected that is neither too cumbersome nor too simplistic, thus achieving the best clustering results. This not only helps improve the efficiency and accuracy of clustering, ensuring that each user's historical spending patterns can be appropriately represented and analyzed, but also avoids wasting computational resources due to too many cluster centers or missing spending patterns due to too few cluster centers.

[0045] After obtaining multiple typical daily spending curves for each user and the typical daily spending curves for user groups at different time periods through clustering algorithms, the real-time daily spending curves of each user on the day they apply for spending are compared with the multiple typical daily spending curves and the typical daily spending curves of user groups. First, the real-time daily spending curves of each user on the day they apply for spending are compared with the multiple typical daily spending curves of the user, resulting in multiple similarity comparison results. The minimum similarity comparison result is selected as the minimum value among the multiple similarity comparison results. This process includes: obtaining the real-time daily spending curves of each user; using the Euclidean distance method to compare the real-time daily spending curves of each user with the multiple typical daily spending curves of the user, resulting in multiple similarity comparison results; and determining the minimum value among the multiple similarity comparison results as the minimum similarity comparison result.

[0046] Specifically, firstly, real-time daily spending data for each user on the day they apply for withdrawal is collected, and a real-time daily spending curve is constructed. This curve reflects the user's actual spending on the day of application, including the amount spent, timestamp, and other information. Next, the Euclidean distance method is used to compare the similarity between the user's real-time daily spending curve and several typical daily spending curves for that user. Euclidean distance is a method for measuring the distance between two points (in this embodiment, two sets of curve data points) in multi-dimensional space, which can intuitively reflect the numerical closeness of two curves. The formula is used... Calculate the Euclidean distance between the user's real-time daily spending curve and each of the user's typical daily spending curves to obtain multiple similarity comparison results, where x 1i This represents the i-th data point on the user's real-time daily spending curve, x. 2i Let represent the i-th data point of the user's typical daily spending curve, N represent the number of data points, and d represent the Euclidean distance between the user's real-time daily spending curve and the user's typical daily spending curve.

[0047] After obtaining multiple similarity comparison results, the minimum similarity result is selected. This minimum similarity result indicates that the Euclidean distance between the user's real-time daily spending curve and a typical daily spending curve is minimized. In other words, the user's real-time daily spending behavior is closest to the user's typical spending behavior over a certain time period. Determining the minimum similarity result helps understand whether the user's spending behavior on that day is consistent with past patterns. If the minimum similarity result is relatively large, it indicates that the user's current spending behavior is abnormal and inconsistent with past patterns, requiring attention.

[0048] Next, the real-time daily spending curve of each user on the day of their withdrawal application is compared with the typical daily spending curves of multiple user groups. In this embodiment, the real-time daily spending curve of each user on the day of their withdrawal application is compared with the typical daily spending curve of the user group to obtain multiple trend comparison results, and the maximum trend comparison result is selected. The maximum trend comparison result is the maximum value among the multiple trend comparison results, including: using the Pearson correlation coefficient to compare the real-time daily spending curve of each user with the typical daily spending curve of the user group to obtain multiple trend comparison results; and determining the maximum value among the multiple trend comparison results as the maximum trend comparison result.

[0049] Specifically, the Pearson correlation coefficient is a statistical indicator that measures the degree of linear correlation between two variables. Its value ranges from -1 to 1, where 1 represents a perfect positive correlation, -1 represents a perfect negative correlation, and 0 represents no correlation. In this embodiment, the formula is used. For each user's application for withdrawal, the real-time daily withdrawal curve on the day of application is compared with the typical daily withdrawal curve of the previously determined user group using Pearson correlation coefficient. Here, X represents each data point on the real-time daily withdrawal curve, Y represents each corresponding data point on the typical daily withdrawal curve of the user group, N represents the number of data points, and P... X,Y This indicates the trend comparison results between the real-time daily spending curve and the typical daily spending curve of the user group. This means that for each real-time daily spending curve, a trend comparison result is calculated between it and the typical daily spending curve of each user group. These trend comparison results reflect the degree of matching between the user's current spending trend and the typical spending trend of the group in history.

[0050] Through the above calculation process, multiple trend comparison results are obtained. The maximum value among these results is identified and determined; this maximum trend comparison result represents the highest degree of correlation between the user's real-time daily spending curve on the day of the withdrawal request and a typical daily spending curve of a certain user group. The core of this calculation process lies in using the Pearson correlation coefficient as a trend comparison tool, which can effectively measure the similarity of two curves in terms of trend, even if the two curves differ numerically. By comparing the user's real-time daily spending curve with the typical daily spending curve of the user group, it is possible to discover whether the user's current spending behavior conforms to the behavioral pattern of their group. In particular, when the maximum trend comparison result is low, it indicates that the user's current spending behavior deviates significantly from the group's historical behavior, i.e., there is a risk of abnormal spending.

[0051] To further enhance the depth and breadth of monitoring consumer loan usage, some embodiments can analyze users' social network data. Specifically, after obtaining user authorization, interaction data on their social networks is collected, including contact frequency and communication content. The interactions related to consumer loan usage on social networks are analyzed to assess the degree of influence of social networks on user spending behavior. The social network impact assessment results are combined with users' historical and real-time spending data for a comprehensive risk assessment. For users negatively affected by social interactions, early risk warnings are issued. Social network analysis helps understand the social factors behind user behavior, improves the comprehensiveness of risk assessment, and, combined with the influence of users' social networks, enables the development of more personalized risk prevention and management strategies, thereby improving the security of fund usage.

[0052] After obtaining the minimum similarity comparison result and the maximum trend comparison result, the minimum similarity comparison result and the maximum trend comparison result are compared with their corresponding preset thresholds to monitor whether the user's spending behavior is abnormal. This includes: determining that the user's spending behavior is abnormal when the minimum similarity comparison result is greater than or equal to the first preset threshold and the maximum trend comparison result is less than or equal to the second preset threshold; determining that the user's spending behavior is not abnormal when the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is greater than the second preset threshold; determining that the user's spending behavior is not abnormal when the minimum similarity comparison result is greater than or equal to the first preset threshold and the maximum trend comparison result is greater than the second preset threshold; and determining that the user's spending behavior is not abnormal when the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is less than or equal to the second preset threshold.

[0053] The above describes the specific logic for determining whether a user's spending behavior is abnormal based on the minimum similarity comparison result and the maximum trend comparison result, combined with preset thresholds. When the minimum similarity comparison result is greater than or equal to the first preset threshold, and the maximum trend comparison result is less than or equal to the second preset threshold, abnormal spending behavior is determined. This indicates that the user's current spending behavior has significantly changed compared to their historical habits, posing an anomaly risk. Furthermore, the user's real-time spending curve has a low correlation with the user group's historical typical spending curve, meaning the user's current spending behavior does not conform to the user group's spending habits. In this case, the user's spending behavior can be judged as an abnormal request and should be monitored and alerted. When the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is greater than the second preset threshold, the user's spending behavior is determined not to be abnormal. This indicates that the user's real-time spending curve is similar to the user's historical typical spending curve, meaning the user's current spending behavior is consistent with their historical habits. Furthermore, the user's real-time spending curve has a high correlation with the user group's historical typical spending curve, meaning the user's current spending behavior conforms to the user group's spending habits. In this case, the user's spending behavior can be determined as a normal request, consistent with their own and the user group's historical spending habits. If the minimum similarity comparison result is greater than or equal to the first preset threshold and the maximum trend comparison result is greater than the second preset threshold, the user's spending behavior is determined to be normal. This indicates that the user's real-time spending curve differs significantly from their historical typical spending curve, meaning the user's current spending behavior has changed compared to their historical habits. However, the user's real-time spending curve has a high correlation with the user group's historical typical spending curve, meaning the user's current spending behavior conforms to the user group's spending habits. In this case, although the user's own spending behavior has changed, it conforms to the user group's habitual trend, indicating that the change is within the expected range; therefore, the user's spending behavior can also be determined as a normal request. If the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is less than or equal to the second preset threshold, the user's spending behavior is determined to be normal. This indicates that the user's real-time spending curve is similar to their historical typical spending curve, meaning the user's current spending behavior is consistent with their historical habits. However, the user's real-time spending curve has a weak correlation with the historical typical spending curve of the user group, meaning the user's current spending behavior differs from the user group's spending habits. In this case, although the user's spending behavior does not conform to the overall trend of the user group, it is consistent with their own historical spending habits; therefore, the user's spending behavior can still be considered a normal request.

[0054] Based on the above logic, multi-dimensional risk assessments can be conducted using users' real-time and historical spending data, combined with group behavior patterns. This allows for the timely detection and early warning of potential abnormal behaviors, thereby effectively managing the risk of consumer loan usage. The advantage of this method lies in considering not only individual user behavior changes but also the consistency of behavior between the user and their group, achieving a comprehensive and forward-looking risk assessment during the consumer loan application stage.

[0055] In some embodiments, deep learning models (such as Long Short-Term Memory networks or convolutional neural networks) can also be used to optimize the prediction and risk assessment of spending behavior. Specifically, user spending data is transformed into a format suitable for input to the deep learning model, such as time-series data. The deep learning model is trained using historical spending data with the goal of predicting user spending behavior patterns and their potential risks. Real-time user spending data is input into the trained model to predict the risk level of user spending behavior in real time. Based on the model's prediction results, the aforementioned preset thresholds are dynamically adjusted to adapt to constantly changing contexts and user behavior patterns. Deep learning models can learn and capture complex patterns in spending behavior, improving prediction accuracy, and the model can update prediction strategies in real time, more flexibly adapting to dynamic changes in the market and user behavior.

[0056] To enable those skilled in the art to better understand the technical solution of this application, the implementation process of the consumer loan fund usage monitoring method of this application will be described in detail below with reference to specific embodiments.

[0057] This embodiment relates to a specific method for monitoring the use of consumer loan funds, such as... Figure 3 As shown, it includes the following steps:

[0058] Step S1: Accessing existing data. Because there may be missing data during the access process, this embodiment uses a formula... The average value of the data for the user at the current time is used to fill in missing data. Where x it This represents missing data for the i-th user at time t. n is the number of days in the selected data sample. xtk is the user's spending limit on day k at time t. z represents the number of times the user's spending limit was used.

[0059] Step S2: Data Preprocessing. User spending data varies greatly depending on their consumption habits. During data clustering, without standardization, the influence of attributes with larger datasets may be amplified while smaller attributes are ignored, leading to inaccurate or even erroneous clustering results. This embodiment uses a deviation standardization algorithm to normalize the data, performing a linear transformation.

[0060] Let dataset X i ={x i1 ,x i2 ....,x in If there are n attributes, then the standardized value is '

[0061] Where, x ij For standardized data, and X i The maximum and minimum values ​​in the cluster are defined. After standardization, all data fall within the range of [0, 1], avoiding the amplification of the influence of certain large-scale values, thus making the clustering results more accurate.

[0062] Step S3: Extraction and Reception of Typical Daily Spending Curves for Users. Because the consumption behavior and habits of the same user vary at different times, to avoid random occurrences, this embodiment first uses the elbow rule to determine the optimal number of clusters, and then uses the K-means++ clustering algorithm to analyze the user's historical consumption data, accurately extracting multiple daily spending curves for the user at different time periods. Thus, multiple typical daily spending curves specific to each user have been extracted.

[0063] Step S4: Measure the user's real-time daily spending curve against the user's typical daily spending curve.

[0064] (1) Use Euclidean distance to measure the similarity between each user’s real-time daily spending curve and each user’s typical daily spending curve, reflecting the differences in user spending data values.

[0065] (2) Record the minimum metric result a.

[0066] Step S5: Extraction of typical daily spending curves for the user group. A clustering algorithm is used to cluster the typical daily spending curves of all users obtained in Step S3. The cluster center curve of the cluster containing a user's typical daily spending curve is determined as the typical daily spending curve of the user group.

[0067] Traditional K-means clustering algorithm is simple to operate and highly scalable, making it suitable for large-scale datasets. However, it suffers from drawbacks such as the need to pre-determine the initial number of cluster centers and the selection of initial centroids. To effectively extract a group of users with similar consumption behaviors to the given user, this embodiment proposes an improved algorithm as follows:

[0068] 1) Use the elbow rule to determine the number of cluster centers k. If the value of k is greater than the number of typical curves for the user, then k is used as the number of cluster centers; if the value of k is less than the number of typical curves for the user, then the number of typical curves for the user is used as the number of cluster centers.

[0069] 2) Use the typical curve of this user calculated in step S3 as the initial centroid.

[0070] 3) Calculate the Euclidean distances from other points in the data (points other than the cluster centers) to each cluster center.

[0071] 4) Recalculate the centroid and repeat step S3 until the centroid no longer changes.

[0072] This method, to a certain extent, avoids the interference of big data on user characteristic behaviors and the trouble of falling into local optimal solutions due to randomly selecting cluster centers.

[0073] Step S6: Measure the real-time daily expenditure curve of this user against the typical daily expenditure curve of the user group to which the user belongs.

[0074] (1) Use the Pearson correlation coefficient to measure the trend of the real-time expenditure curve of the user every day against the typical daily expenditure curve of the user group to which the user belongs, reflecting the difference between the actual expenditure data curve of the user and the daily habitual expenditure trend of the user group to which the user belongs.

[0075] (2) Record the measurement result b with the highest correlation.

[0076] Step S7: Determine whether the user has abnormal behaviors based on the measurement results, so as to monitor and give early warnings.

[0077] Compare the obtained measurement results a and b with the predetermined thresholds A and B. The thresholds A and B need to be obtained through a large amount of actual data for simulation training. In this embodiment, the following situations can be referred to for judgment:

[0078] (1) a≥A and b≤B, indicating that the real-time expenditure curve of the user is quite different from the historical typical expenditure curve of the user, that is, the current expenditure behavior of the user has changed significantly compared with the user's historical behavior habits, and there is an abnormal risk. And the correlation between the real-time expenditure curve of the user and the historical typical expenditure curve of the user group is small, that is, the current expenditure behavior of the user does not conform to the expenditure behavior habits of the user group. In this case, the expenditure behavior of this user can be determined as an abnormal application and should be monitored and given an early warning.

[0079] (2) aB, indicating that the real-time expenditure curve of the user is similar to the historical typical expenditure curve of the user, that is, the current expenditure behavior of the user is consistent with the user's historical behavior habits. And the correlation between the real-time expenditure curve of the user and the historical typical expenditure curve of the user group is high, that is, the current expenditure behavior of the user conforms to the expenditure behavior habits of the user group. In this case, the expenditure behavior of this user can be determined as a normal application, conforming to the historical expenditure habits of the user himself and the user group.

[0080] (3) If a≥A and b>B, it indicates that the user's real-time usage curve is quite different from the user's historical typical usage curve, that is, the user's current usage behavior has changed compared with the user's historical behavior habits. However, the correlation between the user's real-time usage curve and the historical typical usage curve of the user group is relatively high, that is, the user's current usage behavior conforms to the usage behavior habits of the user group. In this case, although the user's own usage behavior has changed but conforms to the habit trend of the user group, it shows that this change is within the expected range. Therefore, the user's usage behavior can also be determined as a normal application.

[0081] (4) If a<A and b≤B, it indicates that the user's real-time usage curve is similar to the user's historical typical usage curve, that is, the user's current usage behavior is consistent with the user's historical behavior habits. However, the correlation between the user's real-time usage curve and the historical typical usage curve of the user group is relatively small, that is, the user's current usage behavior is different from the usage behavior habits of the user group. In this case, although the user's usage behavior does not conform to the overall trend of the user group, it is consistent with the user's own historical usage habits. Therefore, the user's usage behavior can also be determined as a normal application.

[0082] The embodiment of the present application also provides a monitoring device for the use of funds for consumer loans. It should be noted that the monitoring device for the use of funds for consumer loans in the embodiment of the present application can be used to execute the method for monitoring the use of funds for consumer loans provided by the embodiment of the present application. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0083] The following introduces the monitoring device for the use of funds for consumer loans provided by the embodiment of the present application.

[0084] Figure 4 It is a structural block diagram of the monitoring device for the use of funds for consumer loans according to the embodiment of the present application. As Figure 4As shown, the device includes an acquisition unit 10, a first determination unit 20, a similarity comparison unit 30, a trend comparison unit 40, and a comparison unit 50. The acquisition unit acquires user credit limit spending data from multiple users and performs cluster analysis on the user credit limit spending data using a first clustering algorithm to extract multiple typical daily spending curves for each user at different time periods. These time periods include multiple working days within a preset time period and multiple holidays within the preset time period. The determination unit determines the typical daily spending curves for a user group based on the typical daily spending curves of all the users. The similarity comparison unit compares the real-time daily spending curve of each user on the day they apply for credit with the multiple typical daily spending curves of the user. The system performs a similarity comparison, obtaining multiple similarity comparison results and selecting the minimum similarity comparison result, which is the smallest value among the multiple similarity comparison results. The trend comparison unit performs a trend comparison between the real-time daily spending curve of each user and the typical daily spending curve of the user group, obtaining multiple trend comparison results and selecting the maximum trend comparison result, which is the maximum value among the multiple trend comparison results. The comparison unit compares the minimum similarity comparison result and the maximum trend comparison result with the corresponding preset thresholds to monitor whether there are any abnormalities in the spending behavior of the users.

[0085] Specifically, the acquisition unit retrieves user credit limit spending data from relevant databases and performs cluster analysis on this data using a first clustering algorithm. In this embodiment, the first clustering algorithm refers to the K-means++ clustering algorithm, which aims to group user spending behavior and analyze each user's spending habits over different time periods. The cluster analysis assigns data points to different clusters (or groups) based on characteristics of user spending data, such as spending time and amount, with each cluster representing a spending behavior pattern. Through clustering, the consumption habits and patterns of each user over different time periods can be revealed, providing a basis for subsequent risk assessment.

[0086] Through cluster analysis, a curve representing the average spending behavior of users within each cluster is extracted, known as the typical daily spending curve. This curve reflects the average spending pattern of users across different time periods (weekdays and holidays). Specifically, since holiday spending patterns differ significantly from weekday spending patterns, cluster analysis of users' historical spending data identifies clusters with similar spending habits from a large pool of historical data. The cluster center of each cluster is then calculated, and this cluster center forms the typical daily spending curve. Users' spending habits differ between weekdays and holidays. Weekdays can be further subdivided into the beginning, middle, and end of the month, while holidays can be subdivided into weekends, long holidays, short holidays, and shopping festivals. Users' spending data from the past year can be obtained, encompassing various weekday and holiday scenarios. By performing cluster analysis on these user credit limit usage data, typical daily spending curves that represent the user's spending behavior characteristics across different time periods are automatically identified and extracted. These typical daily spending curves are automatically generated based on naturally formed clustering patterns in the user's historical spending data. For example, the cluster analysis results may reveal a curve that particularly matches the user's spending behavior on most weekdays; another curve represents the user's spending behavior on weekends or public holidays; and yet another curve reflects the user's spending behavior during shopping festivals. By using cluster analysis to identify users' typical daily spending curves, it is possible to more accurately capture the user's spending behavior characteristics across different time periods, thereby better identifying potential abnormal spending behavior.

[0087] The determination unit is used to further extract typical daily spending curves for user groups from the typical daily spending curves of all users. The acquisition unit has already performed cluster analysis on the user credit limit spending data for each user, thereby extracting multiple typical daily spending curves for each user over different time periods. These typical daily spending curves reflect the consumption patterns and habits of individual users based on their historical data. Furthermore, to more comprehensively assess whether there are anomalies in user spending behavior, considering only the typical daily spending curves of individual users is insufficient; it is also necessary to consider the behavior of a broader user group. Therefore, it is necessary to extract typical daily spending curves for user groups based on the typical daily spending curves of all users. By dividing users into different groups, further analysis of the typical daily spending curves of all users within each group is conducted to obtain curves that represent the consumption patterns of that group, i.e., the typical daily spending curves of the user group. The typical daily spending curves of each group reflect the common consumption habits of users in that group. In summary, the purpose of extracting typical daily spending curves for user groups is to establish a benchmark, providing a normal consumption pattern for the group to which an individual user belongs, and offering a comparison for subsequent judgment of abnormal user spending behavior.

[0088] The similarity comparison unit compares a user's real-time daily spending curve with each of the user's typical daily spending curves, obtaining multiple similarity comparison results to quantify the degree of similarity between the user's real-time daily spending curve and each of the user's typical daily spending curves. The minimum similarity comparison result is selected from these multiple results. The minimum similarity comparison result means that the difference between the user's real-time daily spending curve and a particular typical daily spending curve of the user is the smallest, that is, the user's real-time spending behavior is closest to the user's typical spending behavior pattern. By selecting the minimum similarity comparison result, a quantitative indicator can be provided for subsequent judgment of abnormal user behavior.

[0089] The trend comparison unit compares each user's real-time daily spending curve with the typical daily spending curves of multiple user groups identified previously. This comparison yields multiple trend comparison results, reflecting the difference in trend between the user's spending behavior on the day they apply for payment and the spending behavior of their group. The unit selects the result with the highest correlation, representing the highest degree of relevance. This means the user's real-time daily spending curve most closely matches the trend of a typical daily spending curve of a particular user group, indicating the most consistent real-time behavior with the historical behavior of their group. Selecting the highest trend comparison result provides a quantitative indicator for subsequent assessment of abnormal user behavior.

[0090] The comparison unit compares the maximum trend comparison result with the corresponding preset threshold, which is obtained through simulation training using a large amount of actual data. This threshold is used to determine whether the trend of user behavior is normal. If the maximum trend comparison result is lower than the preset threshold, it indicates that the user's current spending trend is significantly different from the group behavior, indicating that the user's spending behavior is abnormal. Similarly, the minimum similarity comparison result is compared with the corresponding preset threshold, which is also obtained through simulation training using a large amount of actual data. If the minimum similarity comparison result exceeds the preset threshold, it indicates that the user's current spending behavior deviates significantly from the typical spending pattern in the past, which also indicates that there is an abnormal spending behavior.

[0091] By comparing the minimum similarity comparison results and the maximum trend comparison results with corresponding preset thresholds, it is possible to determine whether there are any anomalies in user spending behavior. The selection of these preset thresholds is crucial; they are set based on extensive historical data and a deep understanding of the user group, aiming to ensure that abnormal spending behavior is captured without misjudging normal spending activities. Using the minimum similarity comparison results and the maximum trend comparison results in conjunction with the preset thresholds, anomalies in user spending behavior can be monitored in real time, potential risks can be identified promptly, and the risk identification capability for fund usage can be improved.

[0092] This embodiment enables immediate risk assessment of users' real-time withdrawal application data on the day of application, overcoming the lag in traditional post-loan fund monitoring in existing technologies. Furthermore, it extracts withdrawal behavior curves for users and user groups based on a large amount of historical user credit limit withdrawal data, making the extracted comparison curves more reliable and allowing for more accurate prediction of whether users' current withdrawals are abnormal. This solves the problem of limited ability to identify abnormal fund usage risks in the consumer loan withdrawal application stage in existing technologies.

[0093] In the specific implementation process, the aforementioned determining unit includes a clustering analysis module and a first determining module. The clustering analysis module uses a second clustering algorithm to perform clustering analysis on the typical daily spending curves of all the aforementioned users, obtaining multiple clusters of typical daily spending curves. The first determining module determines the cluster center curves of the multiple clusters of typical daily spending curves as the typical daily spending curves of the aforementioned user group. The clustering analysis module includes a calculation submodule and an allocation submodule. The calculation submodule, for each of the aforementioned users, uses the typical daily spending curves of each user as the initial centroid, calculates the Euclidean distances between the typical daily spending curves of all other users and the initial centroid, obtaining multiple Euclidean distances for each user. The allocation submodule assigns the typical daily spending curves of each user to the cluster containing the initial centroid corresponding to the smallest Euclidean distance among the multiple Euclidean distances for that user, recalculates the centroid of each cluster, until a preset stopping condition is reached, obtaining multiple clusters of typical daily spending curves for each of the aforementioned users.

[0094] Specifically, the clustering analysis module, after obtaining the typical daily spending curves of all users, uses a second clustering algorithm to cluster all typical daily spending curves. In this embodiment, the second clustering algorithm is the K-means++ clustering algorithm. The purpose of clustering is to group curves with similar consumption behavior patterns together to form multiple clusters of typical daily spending curves. The clustering algorithm automatically groups curves based on the similarity or difference between them. Curves within each group have high similarity in spending behavior, while curves between different groups show significant differences in behavior patterns. The calculation submodule uses the typical daily spending curve of each user as the initial centroid. That is, each user's typical daily spending curve is regarded as a cluster center, which helps ensure that the algorithm can fully consider the historical behavior patterns of each user. For example, when determining the user group to which a user belongs, the user's typical daily spending curve is used as the initial centroid. Then, the Euclidean distance between each typical daily spending curve of other users and the user's initial centroid is calculated to obtain multiple Euclidean distances for each user. Euclidean distance is a method for measuring the spatial distance between two points. In this embodiment, it is used to quantify the differences between curves; the smaller the distance, the more similar the two curves are. The allocation submodule is used to assign each user's typical daily spending curve to the cluster containing the initial centroid corresponding to it with the smallest Euclidean distance. In this way, similar curves will cluster together to form preliminary clusters. This process essentially involves automatically classifying the historical spending behavior of the user group, with each category representing the consumption habits of that user group during a specific period.

[0095] Whenever a user's typical daily spending curve is analyzed and added, it is assigned to the closest cluster based on its similarity to existing curves, and the centroid of the assigned cluster is recalculated. The new centroid can be the average of all curves within the cluster. The entire process (calculating Euclidean distance, assigning curves, and updating centroids) is then repeated until a preset stopping condition is met. In this embodiment, the preset stopping condition can be that the centroid no longer changes, the change in the centroid is less than a certain threshold, or a preset number of iterations has been reached, indicating that the cluster division has stabilized and no further adjustment is needed.

[0096] Through the above process, multiple typical daily spending curve clusters are ultimately obtained, each cluster containing a set of curves with similar spending behaviors. The cluster center curve reflects the typical spending behavior of the user group to which the user belongs within a certain time period. Determining the cluster center curve allows for clear identification of the spending behavior patterns of the user group to which the user belongs in different time periods, providing a benchmark for subsequent comparison of group behavior trends. The first determination module is used to determine the cluster center curve of each user's typical daily spending curve cluster as the typical daily spending curve of that user's user group. Each typical daily spending curve of a user group represents the typical consumption behavior of a type of user within a specific time period. These typical daily spending curves of user groups will be used to compare trends with the real-time daily spending curve of an individual user on the day they apply for spending, thereby assessing whether the user's behavior deviates from the normal pattern of their group and identifying potential abnormal spending behavior.

[0097] In some embodiments, the apparatus includes a second determining unit and a third determining unit. The second determining unit is configured to determine the initial number of cluster centers using the elbow rule before performing cluster analysis on the typical daily spending curves of all the users using a second clustering algorithm, and to compare the initial number of cluster centers with the number of typical daily spending curves of all the users. The third determining unit is configured to determine the initial number of cluster centers as the number of cluster centers if the initial number of cluster centers is greater than or equal to the number of typical daily spending curves of all the users; and to determine the number of typical daily spending curves of all the users as the number of cluster centers if the initial number of cluster centers is less than the number of typical daily spending curves of all the users.

[0098] Specifically, the second determining unit uses the elbow rule to determine a reasonable initial number of cluster centers before performing clustering algorithm analysis. The third determining unit, after obtaining the initial number of cluster centers through the elbow rule, compares it with the number of typical daily spending curves for all users. The number of typical daily spending curves reflects the number of users' historical behavior data clustered into typical curves. When the initial number of cluster centers determined by the elbow rule is greater than or equal to the number of typical daily spending curves for all users, this initial number of cluster centers is used as the total number of cluster centers. This is because a sufficient number of cluster centers ensures that each typical daily spending curve belongs to at least one cluster, avoiding missing clusters. If the initial number of cluster centers is less than the number of typical daily spending curves for all users, the number of typical daily spending curves for all users is used as the total number of cluster centers. This step ensures that the typical daily spending curves of each user are considered, avoiding the neglect of certain user behavior patterns due to insufficient cluster centers.

[0099] By following the above process, before starting cluster analysis, based on the elbow rule and the number of typical daily spending curves for all users, a number of cluster centers can be selected that is neither too cumbersome nor too simplistic, thus achieving the best clustering results. This not only helps improve the efficiency and accuracy of clustering, ensuring that each user's historical spending patterns can be appropriately represented and analyzed, but also avoids wasting computational resources due to too many cluster centers or missing spending patterns due to too few cluster centers.

[0100] In some embodiments, the similarity comparison unit includes an acquisition module, a similarity comparison module, and a second determination module. The acquisition module acquires the real-time daily spending curves of each user; the similarity comparison module uses Euclidean distance to compare the real-time daily spending curves of each user with multiple typical daily spending curves of the user, obtaining multiple similarity comparison results; the second determination module determines the minimum similarity comparison result among the multiple similarity comparison results.

[0101] Specifically, the acquisition module collects real-time daily spending data for each user on the day they apply for a withdrawal and constructs a real-time daily spending curve. This curve reflects the user's actual spending on the day of application, including the amount spent and timestamps. The similarity comparison module uses Euclidean distance to compare the user's real-time daily spending curve with multiple typical daily spending curves of the same user. Euclidean distance is a method for measuring the distance between two points (in this embodiment, two sets of curve data points) in a multi-dimensional space, which can intuitively reflect the numerical similarity between two curves. The Euclidean distance between the user's real-time daily spending curve and each of the user's typical daily spending curves is calculated, resulting in multiple similarity comparison results. The second determination module selects the minimum value among the multiple similarity comparison results, i.e., the minimum similarity comparison result. This minimum similarity comparison result indicates that the Euclidean distance between the user's real-time daily spending curve and a certain typical daily spending curve is the smallest, meaning that the user's real-time daily spending behavior is most similar to the user's typical spending behavior over a certain time period. Determining the minimum similarity comparison result can help understand whether a user's spending behavior on a given day is consistent with past behavior patterns. If the minimum similarity comparison result is relatively large, it indicates that the user's current spending behavior is abnormal and inconsistent with past behavior patterns, which requires attention.

[0102] The aforementioned trend comparison unit includes a trend comparison module and a third determination module. The trend comparison module uses the Pearson correlation coefficient to compare the real-time daily spending curves of each user with the typical daily spending curves of the user group, obtaining multiple trend comparison results. The third determination module determines the maximum value among the multiple trend comparison results as the maximum trend comparison result.

[0103] Specifically, the comparison module is used to calculate the Pearson correlation coefficient between the real-time daily spending curve of each user on the day of application and the typical daily spending curve of the previously determined user group. This means that for each real-time daily spending curve, a trend comparison result between it and the typical daily spending curve of each user group will be calculated. These trend comparison results reflect the degree of matching between the user's current spending trend and the typical spending trend of the group in history.

[0104] The third determination module is used to identify the maximum value among these trend comparison results, namely the maximum trend comparison result. This maximum trend comparison result represents the highest degree of correlation between the user's real-time daily spending curve on the day of the withdrawal request and the typical daily spending curve of a certain user group. The core of the above calculation process lies in using the Pearson correlation coefficient as a trend comparison tool, which can effectively measure the similarity of the two curves in terms of trends, even if the two curves differ in numerical value. By comparing the trends of the user's real-time daily spending curve and the typical daily spending curve of the user group, it is possible to discover whether the user's current spending behavior conforms to the behavioral pattern of their group. In particular, when the maximum trend comparison result is low, it indicates that the user's current spending behavior deviates significantly from the group's historical behavior, i.e., there is a risk of abnormal spending.

[0105] The comparison unit includes a fourth determining module, a fifth determining module, a sixth determining module, and a seventh determining module. Specifically, the fourth determining module, after obtaining the minimum similarity comparison result and the maximum trend comparison result, determines that the user's spending behavior is abnormal if the minimum similarity comparison result is greater than or equal to a first preset threshold and the maximum trend comparison result is less than or equal to a second preset threshold. The fifth determining module determines that the user's spending behavior is not abnormal if the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is greater than the second preset threshold. The sixth determining module determines that the user's spending behavior is not abnormal if the minimum similarity comparison result is greater than or equal to the first preset threshold and the maximum trend comparison result is greater than the second preset threshold. The seventh determining module determines that the user's spending behavior is not abnormal if the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is less than or equal to the second preset threshold.

[0106] Specifically, when the minimum similarity comparison result is greater than or equal to the first preset threshold, and the maximum trend comparison result is less than or equal to the second preset threshold, abnormal user behavior is determined. This indicates that the user's current spending behavior has significantly changed compared to the user's historical habits, posing an anomaly risk. Furthermore, the user's real-time spending curve has a low correlation with the user group's historical typical spending curve, meaning the user's current spending behavior does not conform to the user group's spending habits. In this case, the user's spending behavior can be judged as an abnormal request and should be monitored and alerted. Conversely, when the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is greater than the second preset threshold, the user's spending behavior is determined to be normal. This indicates that the user's real-time spending curve is similar to the user's historical typical spending curve, meaning the user's current spending behavior is consistent with the user's historical habits. Furthermore, the user's real-time spending curve has a high correlation with the user group's historical typical spending curve, meaning the user's current spending behavior conforms to the user group's spending habits. In this case, the user's spending behavior can be judged as a normal request, consistent with the user's own and the user group's historical spending habits. If the minimum similarity comparison result is greater than or equal to the first preset threshold and the maximum trend comparison result is greater than the second preset threshold, the user's spending behavior is determined to be normal. This indicates that the user's real-time spending curve differs significantly from the user's historical typical spending curve, meaning the user's current spending behavior has changed compared to their historical habits. However, the user's real-time spending curve has a high correlation with the historical typical spending curve of the user group, meaning the user's current spending behavior conforms to the user group's spending habits. In this case, although the user's spending behavior has changed, it conforms to the user group's habitual trend, indicating that the change is within the expected range, and therefore the user's spending behavior can be judged as a normal request. If the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is less than or equal to the second preset threshold, the user's spending behavior is determined to be normal. This indicates that the user's real-time spending curve is similar to the user's historical typical spending curve, meaning the user's current spending behavior is consistent with their historical habits. However, the user's real-time spending curve has a low correlation with the historical typical spending curve of the user group, meaning the user's current spending behavior differs from the user group's spending habits. In this case, although the user's spending behavior does not conform to the overall trend of the user group, it is consistent with their own historical spending habits. Therefore, the user's spending behavior can also be judged as a normal application.

[0107] Based on the above logic, multi-dimensional risk assessments can be conducted using users' real-time and historical spending data, combined with group behavior patterns. This allows for the timely detection and early warning of potential abnormal behaviors, thereby effectively managing the risk of consumer loan usage. The advantage of this method lies in considering not only individual user behavior changes but also the consistency of behavior between the user and their group, achieving a comprehensive and forward-looking risk assessment during the consumer loan application stage.

[0108] The aforementioned consumer loan fund usage monitoring device includes a processor and a memory. The acquisition unit, first determination unit, similarity comparison unit, trend comparison unit, and comparison unit are all stored as program units in the memory. The processor executes these program units stored in the memory to achieve their respective functions. All of the above modules reside in the same processor; alternatively, the modules may be located in different processors in any combination.

[0109] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0110] This invention provides a computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the consumer loan fund usage monitoring method.

[0111] This invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the above-mentioned method for monitoring the use of consumer loan funds.

[0112] This application also provides a computer program product that, when executed on a data processing device, is adapted to perform the steps of initializing the above-described method for monitoring the use of funds in consumer loans.

[0113] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0114] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0115] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0116] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0118] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0119] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0120] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0121] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0122] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for monitoring the use of consumer loan funds, characterized in that, include: The system acquires user credit limit spending data from multiple users and uses a first clustering algorithm to perform cluster analysis on the user credit limit spending data from multiple users, extracting multiple typical daily spending curves for each user in different time periods. The time periods include a time period consisting of multiple working days within a preset time period and a time period consisting of multiple holidays within the preset time period. Based on the typical daily spending curves of all the users, determine the typical daily spending curves of the user group; The real-time daily spending curve of each user on the day of application is compared with the user's multiple typical daily spending curves to obtain multiple similarity comparison results. The minimum similarity comparison result is selected as the minimum value among the multiple similarity comparison results. The real-time daily spending curve of each user is compared with the typical daily spending curve of the user group to obtain multiple trend comparison results, and the maximum trend comparison result is selected. The maximum trend comparison result is the maximum value among the multiple trend comparison results. The minimum similarity comparison result and the maximum trend comparison result are compared with their respective preset thresholds to monitor whether there are any abnormalities in the user's spending behavior.

2. The method according to claim 1, characterized in that, Based on the typical daily spending curves of all the aforementioned users, determine the typical daily spending curves for the user group, including: The second clustering algorithm is used to perform cluster analysis on the typical daily spending curves of all the users to obtain multiple clusters of typical daily spending curves; The cluster center curve of the multiple typical daily spending curve clusters is determined as the typical daily spending curve of the user group.

3. The method according to claim 2, characterized in that, The second clustering algorithm was used to perform cluster analysis on the typical daily spending curves of all the users, resulting in multiple clusters of typical daily spending curves, including: Using the second clustering algorithm, for each user, the typical daily spending curve of each user is used as the initial centroid, and the Euclidean distance between the typical daily spending curve of all other users except the user and the initial centroid is calculated to obtain multiple Euclidean distances for each user. Each user's typical daily expenditure curve is assigned to the cluster containing the initial centroid of the smallest Euclidean distance among the user's multiple Euclidean distances. The centroid of each cluster is recalculated until a preset stopping condition is met, thus obtaining multiple clusters of typical daily expenditure curves for each user.

4. The method according to claim 2, characterized in that, Before performing cluster analysis on the typical daily spending curves of all the users using the second clustering algorithm, the method includes: The elbow rule is used to determine the initial number of cluster centers, and the initial number of cluster centers is compared with the number of typical daily expenditure curves for all the users. If the initial number of cluster centers is greater than or equal to the number of typical daily spending curves for all users, the initial number of cluster centers is determined as the number of cluster centers. If the initial number of cluster centers is less than the number of typical daily spending curves for all users, the number of typical daily spending curves for all users is determined as the number of cluster centers.

5. The method according to claim 1, characterized in that, The real-time daily spending curve of each user on the day of their spending request is compared with multiple typical daily spending curves of the user to obtain multiple similarity comparison results. The minimum similarity comparison result is selected, which is the minimum value among the multiple similarity comparison results, including: Obtain the real-time daily spending curves of each user; The Euclidean distance method is used to compare the real-time daily spending curve of each user with the user's multiple typical daily spending curves to obtain multiple similarity comparison results; The minimum value among the multiple similarity comparison results is determined as the minimum similarity comparison result.

6. The method according to claim 1, characterized in that, The real-time daily spending curves of each user are compared with the typical daily spending curves of the user group to obtain multiple trend comparison results. The maximum trend comparison result is selected, which is the maximum value among the multiple trend comparison results, including: The real-time daily spending curves of each user are compared with the typical daily spending curves of the user group using the Pearson correlation coefficient to obtain multiple trend comparison results. The maximum value among the multiple trend comparison results is determined as the maximum trend comparison result.

7. The method according to claim 1, characterized in that, The minimum similarity comparison result and the maximum trend comparison result are compared with their corresponding preset thresholds to monitor whether the user's spending behavior is abnormal, including: If the minimum similarity comparison result is greater than or equal to a first preset threshold and the maximum trend comparison result is less than or equal to a second preset threshold, it is determined that the user's spending behavior is abnormal. If the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is greater than the second preset threshold, it is determined that the user's spending behavior is not abnormal. If the minimum similarity comparison result is greater than or equal to the first preset threshold and the maximum trend comparison result is greater than the second preset threshold, it is determined that the user's spending behavior is not abnormal. If the minimum similarity comparison result is less than the first preset threshold and the maximum trend comparison result is less than or equal to the second preset threshold, it is determined that the user's spending behavior is not abnormal.

8. A device for monitoring the use of consumer loan funds, characterized in that, include: The acquisition unit is used to acquire user credit limit spending data of multiple users, and to perform cluster analysis on the multiple user credit limit spending data using a first clustering algorithm to extract multiple typical daily spending curves of each user in different time periods. The time period includes a time period consisting of multiple working days within a preset time period and a time period consisting of multiple holidays within the preset time period. The first determining unit is used to determine the typical daily spending curve of the user group based on the typical daily spending curve of all the users. The similarity comparison unit is used to compare the real-time daily spending curve of each user on the day of application with multiple typical daily spending curves of the user, obtain multiple similarity comparison results, and select the minimum similarity comparison result, which is the minimum value among the multiple similarity comparison results. The trend comparison unit is used to compare the real-time daily spending curve of each user with the typical daily spending curve of the user group to obtain multiple trend comparison results, and select the maximum trend comparison result, wherein the maximum trend comparison result is the maximum value among the multiple trend comparison results. The comparison unit is used to compare the minimum similarity comparison result and the maximum trend comparison result with the corresponding preset thresholds to monitor whether there are any abnormalities in the user's spending behavior.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the consumer loan fund usage monitoring method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including a method for monitoring the use of funds for a consumer loan as described in any one of claims 1 to 7.

Images