Auto finance fraud detection method, system and device based on deep learning

Abstract

The present invention discloses a deep learning-based automobile finance fraud detection method, system, and device, relating to the field of data processing technology. The method comprises: obtaining a transaction subject, obtaining historical financial transaction information and historical detection time periods based on the transaction subject, and obtaining the number of initiations, initiation time points, and initiation amounts; generating a time series, obtaining time interval data, obtaining a first trend indicator, and obtaining a first parameter; generating an amount series, obtaining amount difference data, obtaining a second trend indicator, and obtaining a second parameter; obtaining an initial probability threshold and an adjustment parameter, modifying the initial probability threshold based on the adjustment parameter and obtaining a target probability threshold; obtaining a target fraud probability based on a probability model, the first parameter, and the second parameter; and obtaining an automobile finance fraud detection result based on the target fraud probability and the target probability threshold. The present invention has the advantages of dynamic adjustment, high detection efficiency, and good detection effect.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, system, and device for detecting automobile financial fraud based on deep learning. Background Art

[0002] With the rapid development of the auto finance industry, auto finance transactions are becoming increasingly frequent, providing consumers and businesses with convenient financing and payment methods. However, the attendant risk of financial fraud has also become increasingly prominent, becoming a key factor hindering the healthy development of the auto finance industry. Financial fraud not only causes direct economic losses to financial institutions but also can undermine consumer trust and impact market stability. Meanwhile, in recent years, deep learning technology has made significant progress in various fields. Its advantages in feature extraction, pattern recognition, and data mining have provided new insights into financial fraud detection. Deep learning can automatically extract useful feature representations from large amounts of transaction data without manual intervention, significantly improving the efficiency and accuracy of data acquisition. However, applying deep learning to auto finance fraud detection still faces many challenges.

[0003] On the one hand, auto finance transaction data extracted through deep learning features time series, multidimensionality, and heterogeneity. Effectively leveraging these characteristics to aid subsequent data processing is crucial for financial fraud detection. On the other hand, existing fraud detection methods often use fixed probability thresholds or rules to determine whether a transaction presents fraud risk. This approach ignores the variability between different transaction entities and scenarios, limiting the accuracy and reliability of fraud detection. Summary of the Invention

[0004] In response to the deficiencies in the prior art, the present invention provides a method, system, and device for detecting automobile financial fraud based on deep learning.

[0005] A deep learning-based automobile finance fraud detection method includes: obtaining a transaction subject associated with automobile finance transaction data, obtaining historical financial transaction information and a historical detection time period based on the transaction subject, and obtaining the number of initiations, initiation time points, and initiation amounts of automobile finance transaction data associated with the transaction subject within the historical detection time period based on the deep learning and the historical financial transaction information; sorting multiple initiation time points in chronological order to generate a time series, obtaining time interval data between adjacent initiation time points based on the time series, obtaining a first trend indicator based on the multiple time interval data, and obtaining a first parameter based on a first calculation model and the first trend indicator; sorting multiple initiation amounts in chronological order to generate an amount series, obtaining amount difference data between adjacent initiation amounts based on the amount series, obtaining a second trend indicator based on the multiple amount difference data, and obtaining a second parameter based on a second calculation model and the second trend indicator; obtaining an initial probability threshold based on the number of initiations, obtaining an adjustment parameter based on a correction model, the first trend indicator, and the second trend indicator, correcting the initial probability threshold based on the adjustment parameter to obtain a target probability threshold, obtaining a target fraud probability based on the probability model, the first parameter, and the second parameter, and obtaining an automobile finance fraud detection result based on the target fraud probability and the target probability threshold.

[0006] Optionally, obtaining historical financial transaction information and historical detection time periods based on the transaction subject includes: identifying the identity information of the transaction subject; using the identity information to access a database or data storage system to retrieve and extract historical financial transaction records associated with the transaction subject; and determining the historical detection time period based on the identity information.

[0007] Optionally, obtaining automobile financial fraud detection results based on the target fraud probability and the target probability threshold includes: determining whether the target fraud probability exceeds the target probability threshold; if it exceeds, determining that the transaction subject has a fraud risk in the automobile financial transaction data; if it does not exceed, determining that the transaction subject has no fraud risk in the automobile financial transactions within the historical detection time period, and obtaining the risk interval based on the target probability threshold and the target fraud probability.

[0008] Optionally, the first trend indicator is obtained based on multiple time interval data as follows: Among them, TR1 is the first trend indicator, ΔT i is the time interval data between the i-th adjacent initiation time points in the time series, ΔT i+1 is the time interval data between the i+1th adjacent initiation time points in the time series, n is the number of time interval data in the time series, and α is the impact degree.

[0009] Optionally, the first calculation model in obtaining the first parameter based on the first calculation model and the first trend indicator is expressed as: Among them, P1 is the first parameter, TR1 is the first trend indicator, ΔT i is the time interval data between the i-th adjacent initiation time points in the time series, and n is the number of time interval data in the time series.

[0010] Optionally, the correction model in obtaining the adjustment parameter based on the correction model, the first trend indicator, and the second trend indicator is expressed as: Among them, β is the adjustment parameter, t is the standard value of the parameter, TR1 is the first trend indicator, and TR2 is the second trend indicator.

[0011] Optionally, the probability model in obtaining the target fraud probability according to the probability model, the first parameter, and the second parameter is expressed as: Among them, P t is the target fraud probability, P1 is the first parameter, and P2 is the second parameter.

[0012] A deep learning-based automobile financial fraud detection system is also provided. The system is used to implement any one of the deep learning-based automobile financial fraud detection methods. The system includes: an acquisition module for acquiring a transaction subject associated with automobile financial transaction data, and acquiring historical financial transaction information and a historical detection time period based on the transaction subject, and acquiring the number of initiation times, initiation time points, and initiation amounts of automobile financial transaction data associated with the transaction subject within the historical detection time period based on deep learning and the historical financial transaction information; a first calculation module for sorting multiple initiation time points in chronological order and generating a time series, acquiring time interval data between adjacent initiation time points based on the time series, acquiring a first trend indicator based on the multiple time interval data, and obtaining a first trend indicator based on the multiple time interval data. A first parameter is obtained based on a first calculation model and a first trend indicator; a second calculation module is used to sort multiple initiation amounts in time series order and generate an amount sequence, obtain amount difference data between adjacent initiation amounts based on the amount sequence, obtain a second trend indicator based on multiple amount difference data, and obtain a second parameter based on the second calculation model and the second trend indicator; a fraud detection module is used to obtain an initial probability threshold based on the number of initiations, obtain an adjustment parameter based on the correction model, the first trend indicator and the second trend indicator, correct the initial probability threshold based on the adjustment parameter and obtain a target probability threshold, obtain a target fraud probability based on the probability model, the first parameter and the second parameter, and obtain an automobile finance fraud detection result based on the target fraud probability and the target probability threshold.

[0013] An electronic device is also provided, comprising: a memory storing a computer program; and a processor for executing the computer program in the memory to implement the above-mentioned deep learning-based automobile finance fraud detection method.

[0014] Also provided is a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned deep learning-based automobile finance fraud detection method.

[0015] The beneficial effects of the present invention are embodied in:

[0016] The deep learning-based auto finance fraud detection method utilizes deep learning technology to automatically and accurately extract features closely related to fraud detection, such as the number of transactions initiated, the time of initiation, and the amount of transactions, from large amounts of transaction data, improving the efficiency and accuracy of data processing. Furthermore, the solution conducts in-depth time series analysis and trend mining on the time of initiation and the amount of transactions. By calculating time interval data, amount difference data, and corresponding trend indicators, it quantifies the behavioral trends of transaction entities in terms of time and amount, providing richer and more detailed information for subsequent fraud detection. Furthermore, a dynamically adjusted probability threshold mechanism is introduced. Based on the historical behavioral characteristics of transaction entities and current transaction status, adjustment parameters are calculated through a modified model to modify the initial probability threshold, resulting in a target probability threshold that better matches actual transaction conditions. This effectively addresses the problem in traditional fraud detection methods where fixed probability thresholds or rules ignore the differences between different transaction entities and transaction scenarios, significantly improving the accuracy and reliability of fraud detection. Furthermore, the target fraud probability is calculated through a probability model and compared with the target probability threshold to obtain accurate fraud detection results, providing strong support for risk management and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings required for the specific embodiments or the description of the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. Elements or parts in the drawings are not necessarily drawn to scale.

[0018] Figure 1 This is a schematic diagram of the steps of the deep learning-based automobile finance fraud detection method of the present invention;

[0019] Figure 2 Schematic diagram of some steps S1 in the deep learning-based automobile finance fraud detection method of the present invention;

[0020] Figure 3 This is a schematic diagram of some steps in S4 of the deep learning-based automobile finance fraud detection method of the present invention;

[0021] Figure 4 The present invention is a block diagram of an electronic device according to an embodiment of the present invention.

[0022] Reference numerals:

[0023] 700 - electronic device, 701 - processor, 702 - memory, 703 - multimedia component, 704 - I / O interface, 705 - communication component. DETAILED DESCRIPTION

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0025] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0026] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. In addition, the terms "first," "second," etc. are used only to distinguish the descriptions and are not to be understood as indicating or implying relative importance.

[0027] like Figure 1 As shown, a deep learning-based automobile finance fraud detection method is provided, including:

[0028] S1. Obtaining a transaction subject associated with the automobile financial transaction data, and obtaining historical financial transaction information and a historical detection time period based on the transaction subject. Furthermore, based on deep learning and the historical financial transaction information, obtaining the number of initiations, initiation time points, and initiation amounts of automobile financial transaction data associated with the transaction subject within the historical detection time period;

[0029] S2. Sort the multiple initiation time points in chronological order to generate a time series, obtain time interval data between adjacent initiation time points based on the time series, obtain a first trend indicator based on the multiple time interval data, and obtain a first parameter based on the first calculation model and the first trend indicator;

[0030] S3. Sort the multiple initiated amounts in chronological order to generate an amount sequence, obtain amount difference data between adjacent initiated amounts based on the amount sequence, obtain a second trend indicator based on the multiple amount difference data, and obtain a second parameter based on the second calculation model and the second trend indicator;

[0031] S4. Obtain an initial probability threshold based on the number of initiations, obtain an adjustment parameter based on the correction model, the first trend indicator, and the second trend indicator, correct the initial probability threshold based on the adjustment parameter and obtain a target probability threshold, obtain a target fraud probability based on the probability model, the first parameter, and the second parameter, and obtain an automobile finance fraud detection result based on the target fraud probability and the target probability threshold.

[0032] In this embodiment, it should be noted that S1 first involves obtaining the transaction subject associated with the current auto finance transaction data. This process comprehensively collects and identifies the specific individual or entity conducting the transaction, which could be an individual consumer, a business, or any other legal entity involved in the auto finance transaction. Next, based on the unique identifier of the transaction subject, the transaction subject's historical financial transaction information is deeply mined. This information includes, but is not limited to, past transaction records, trading habits, and credit status, providing rich context for subsequent fraud detection. A historical detection time period is also determined. This time period is set based on the characteristics and needs of the transaction subject to ensure that the analyzed data is both representative and reflects the transaction subject's latest behavioral patterns. In this step, the application of deep learning technology enables the efficient extraction of features closely related to the transaction subject and crucial for fraud detection from massive amounts of data. For example, the feature extraction layer within the deep learning model extracts features from pre-processed transaction records, extracting key features relevant to fraud detection, such as the number of transactions initiated, the time of initiation, and the amount of the transaction.

[0033] Let's take an example to illustrate step S1: Suppose a transaction entity called "XX Auto Finance Company" is applying for a new auto loan. In step S1, "XX Auto Finance Company" is first identified and confirmed as the transaction entity. Subsequently, by accessing an internal database or external data warehouse, all auto finance transaction records of the company within the past year (based on the historical detection period obtained for the transaction entity being one year) are retrieved and extracted. These records may include the number of loan applications initiated by the company, the submission time of each application (i.e., the initiation time point), and the loan amount applied for (i.e., the initiation amount). The collection and analysis of this information can provide a comprehensive and accurate data foundation for subsequent fraud detection.

[0034] S2 is primarily responsible for performing time series analysis on the extracted initiation time points. First, multiple initiation time points are sorted chronologically to generate a clear time series. This time series reflects the temporal distribution and frequency of auto finance transactions conducted by the transaction entity during the historical monitoring period. Next, based on this time series, the time interval data between adjacent initiation time points is calculated. This time interval data can reveal regularities and anomalies in the transaction entity's transactions, such as whether transactions become increasingly concentrated over time or whether there are sudden changes in transaction frequency. Based on this time interval data, a primary trend indicator is calculated to quantify the transaction entity's transaction behavior trends over time. Finally, a primary calculation model is used to combine this primary trend indicator to calculate a primary parameter. This primary parameter serves as an important basis for subsequently adjusting the probability threshold and calculating the target fraud probability, helping to more accurately assess the transaction entity's fraud risk.

[0035] To illustrate step S2, let's use an example: Suppose "XX Auto Finance Company" initiated multiple auto loan applications over the past year, and the submission time of each application was recorded. In step S2, these submission times are first sorted chronologically to generate a time series. Next, the time intervals between adjacent submissions are calculated. For example, the interval between the first and second applications is 30 days, the interval between the second and third applications is 20 days, and so on. Next, based on these time interval data, a first trend indicator is calculated, such as the average time interval and the variance of the time interval. Finally, a first parameter is calculated using the first calculation model and this first trend indicator. This first parameter reflects the transaction behavior trends of "XX Auto Finance Company" over time, providing strong support for subsequent fraud detection.

[0036] In S3, we focus on sequence analysis and trend mining of initiation amounts. First, we sort multiple initiation amounts into time series order to generate an amount sequence. This amount sequence clearly illustrates the distribution and changing trends of the amounts of auto finance transactions conducted by the transacting entity during the historical monitoring period. Next, based on this amount sequence, we calculate the difference between adjacent initiation amounts. This difference data can reveal regularities and unusual fluctuations in transaction amounts, such as whether the transaction amount increases or decreases gradually or whether there are sudden jumps. Then, based on this amount difference data, we calculate a secondary trend indicator, which quantifies the transaction behavior trends of the transacting entity in terms of amount. Finally, we use a secondary calculation model, combined with this secondary trend indicator, to calculate a secondary parameter. This secondary parameter serves as an important reference for subsequently adjusting the probability threshold and calculating the target fraud probability, facilitating a more comprehensive assessment of the transacting entity's fraud risk.

[0037] To illustrate step S3, let's use an example: Suppose "XX Auto Finance Company" initiated multiple auto loan applications over the past year, and the loan amount for each application was recorded. In step S3, these loan amounts are first sorted chronologically to generate a series of amounts. Next, the difference between consecutive loan amounts is calculated. For example, if the difference between the first and second applications is 500,000 yuan, the difference between the second and third applications is -200,000 yuan (indicating a decrease in amount), and so on. Next, based on these amount difference data, a secondary trend indicator is calculated, such as the average amount difference or the variance of the amount difference. Finally, a secondary calculation model is used, combined with this secondary trend indicator, to calculate a secondary parameter. This secondary parameter reflects the transaction behavior trends of "XX Auto Finance Company" in terms of amount, providing more comprehensive data support for subsequent fraud detection.

[0038] In S4, the process determines whether a transaction carries a fraud risk based on the features extracted and parameters calculated in the previous steps. First, an initial probability threshold is obtained based on the number of initiations. This threshold, set based on historical data and security requirements, is used to determine whether fraud exists. Then, a correction model is used to calculate an adjustment parameter, combining the first trend indicator (reflecting transaction behavior trends over time) and the second trend indicator (reflecting transaction behavior trends over amount). This adjustment parameter modifies the initial probability threshold, taking into account behavioral changes over time and amount, making the probability threshold more consistent with the current transaction. Next, a target fraud probability is calculated using the probability model, combining the first parameter (calculated over time) and the second parameter (calculated over amount). This probability reflects the likelihood of fraud risk in the current transaction. Finally, the target fraud probability is compared with the target probability threshold. If the target fraud probability exceeds the target probability threshold, the transaction entity is deemed to carry a fraud risk in the auto finance transaction data. Otherwise, the transaction entity is deemed to have no fraud risk in its auto finance transactions during the historical detection period. A risk range is then calculated based on the target probability threshold and the target fraud probability for further risk assessment and management.

[0039] To illustrate step S4, let's use an example: Suppose that the auto loan application data from "XX Auto Finance Company" over the past year has been processed through steps S1 through S3, resulting in the number of initiations, the first parameter for the time dimension, and the second parameter for the amount dimension. In step S4, an initial probability threshold is first determined based on the number of initiations, for example, set to 0.5 (indicating that a fraud risk exists when the fraud probability exceeds 0.5). Then, using the correction model, combined with the first and second trend indicators, an adjustment parameter, for example, 0.8, is calculated. This adjustment parameter makes the probability threshold more consistent with the actual transaction data of "XX Auto Finance Company." The initial probability threshold is then modified based on the adjustment parameter to obtain a target probability threshold, for example, 0.4 by multiplying 0.5 by 0.8. Next, using the probability model, combined with the first and second parameters, a target fraud probability, for example, 0.6, is calculated. Finally, the target fraud probability of 0.6 is compared with the target probability threshold of 0.4. Since 0.6 is greater than 0.4, it is determined that "XX Auto Finance Company" faces a fraud risk in the current transaction, requiring further investigation and action.

[0040] In summary, the entire deep learning-based auto finance fraud detection method, firstly, through the application of deep learning technology, can automatically and accurately extract features closely related to fraud detection from large amounts of transaction data, such as the number of transactions initiated, the time of initiation, and the amount of initiation, thereby improving the efficiency and accuracy of data processing. Furthermore, the scheme conducts in-depth time series analysis and trend mining on the initiation time and amount. By calculating time interval data, amount difference data, and corresponding trend indicators, it quantifies the behavioral trends of transaction entities in the time and amount dimensions, providing richer and more detailed information for subsequent fraud detection. Furthermore, a dynamically adjusted probability threshold mechanism is introduced. Based on the historical behavioral characteristics of the transaction entity and the current transaction situation, the adjustment parameters are calculated through the correction model, and the initial probability threshold is modified to obtain a target probability threshold that is more consistent with the actual transaction situation. This effectively solves the problem that fixed probability thresholds or rules in traditional fraud detection methods ignore the differences between different transaction entities and transaction scenarios, and significantly improves the accuracy and reliability of fraud detection. Furthermore, the target fraud probability is calculated through the probability model and compared with the target probability threshold to obtain accurate fraud detection results, providing strong support for risk management and decision-making.

[0041] like Figure 2 As shown, in one embodiment, obtaining historical financial transaction information and historical detection time periods according to the transaction subject in S1 includes:

[0042] S11. Identify the identity information of the transaction subject;

[0043] S12. Using the identity information to access a database or data storage system, retrieve and extract historical financial transaction records associated with the transaction subject;

[0044] S13. Determine a historical detection time period based on the identity identification information.

[0045] In this embodiment, it should be noted that in S11, the identity information of the transaction subject must first be identified. This step is the basis for ensuring that historical financial transaction records related to a specific transaction subject can be accurately extracted later. Identity information includes but is not limited to the name of the transaction subject, ID number (for individual consumers), business registration number (for enterprises) or any other information that can uniquely identify the transaction subject. Through these identity information, it is possible to clearly identify the subject of the current transaction, thereby providing accurate positioning for subsequent data extraction and analysis. For example, if the transaction subject is a company called "XX Auto Finance Company", then its corporate registration number is its unique identity information, and this registration number will be used to identify and confirm the transaction subject.

[0046] In S12, the identity information identified in step S11 is used to access a database or data warehouse to retrieve and extract historical financial transaction records associated with the transaction subject. This step is the process of obtaining key information such as the transaction subject's past transaction behavior, trading habits, and credit status. Databases or data warehouses store a large amount of financial transaction data. Using identity information as a search criterion can quickly locate data records related to a specific transaction subject. For example, for "XX Auto Finance Company," the company registration number is used to access the database and extract all of the company's auto finance transaction records from the past year, including detailed information such as the number of loan applications, the submission time of each application, and the application amount.

[0047] In S13, the historical detection time period is determined based on the identity identification information. This step is to ensure that the analyzed data is both representative and able to reflect the latest behavior patterns of the transaction subject. The determination of the historical detection time period can be set according to the characteristics and needs of the transaction subject. For example, different time periods such as the past year, six months or three months can be selected. In S13, the most appropriate historical detection time period will be determined by comprehensively considering factors such as the transaction subject's historical transaction records, transaction frequency, and current transaction conditions. For example, for "XX Auto Finance Company", the past year may be selected as the historical detection time period based on its past transaction records and the current market environment to ensure that the analyzed data can accurately reflect the company's latest transaction behavior patterns.

[0048] like Figure 3 As shown, in one embodiment, obtaining the automobile finance fraud detection result according to the target fraud probability and the target probability threshold in S4 includes:

[0049] S41. Determine whether the target fraud probability exceeds a target probability threshold;

[0050] S42. If the number exceeds the limit, it is determined that the transaction subject has a fraud risk in the automobile finance transaction data;

[0051] S43. If it does not exceed, it is determined that there is no fraud risk in the automobile financial transactions of the transaction subject during the historical detection period, and a risk interval is obtained according to the target probability threshold and the target fraud probability.

[0052] In this embodiment, it should be noted that in S41, the target fraud probability calculated using the probability model is directly compared with the target probability threshold adjusted using the modified model. The comparison result triggers the subsequent fraud risk determination process. For example, if the target fraud probability is 0.6 and the target probability threshold is 0.4, since 0.6 is greater than 0.4, this discrepancy is identified, and preparations are made for S42, which determines that the transaction subject has a fraud risk.

[0053] In S42, if the results of S41 indicate that the target fraud probability exceeds the target probability threshold, step S42 is initiated, formally determining that the transaction entity presents a fraud risk in the auto finance transaction data. This determination is based on the data processing and calculation results of all previous steps and is the result of an in-depth analysis of the transaction entity's behavior patterns. Continuing with the above example, since the target fraud probability of 0.6 for "XX Auto Finance Company" exceeds the target probability threshold of 0.4, a fraud risk report is generated, indicating that the company presents a fraud risk and requires further investigation.

[0054] In S43, if the result of S41 indicates that the target fraud probability does not exceed the target probability threshold, S43 is executed, determining that the transaction subject's auto finance transactions within the historical detection period are free of fraud risk. To further refine risk management, a risk interval is calculated based on the difference between the target probability threshold and the target fraud probability. This risk interval reflects the relative risk level of the transaction subject's current transaction behavior. Even if the transaction does not pose a fraud risk, the risk interval can provide additional reference information for risk managers. For example, if the target fraud probability is 0.3 and the target probability threshold is 0.4, not only will "XX Auto Finance Company" be determined to pose no fraud risk in the current transaction, but a risk interval, such as 0.1 (i.e., 0.4-0.3), will also be calculated, indicating that the company's transaction behavior is still within a safe distance from the fraud risk threshold. However, risk managers still need to monitor risk fluctuations within this interval.

[0055] In one embodiment, obtaining the first trend indicator according to the multiple time interval data in S2 is expressed as:

[0056] in,

[0057] TR1 is the first trend indicator, ΔT i is the time interval data between the i-th adjacent initiation time points in the time series, ΔT i+1 is the time interval data between the i+1th adjacent initiation time points in the time series, n is the number of time interval data in the time series, and α is the impact degree.

[0058] In this embodiment, it should be noted that the sign function sgn(*) is used to determine the direction of change of the time interval; if the variable is greater than 0, sgn(*) = 1; if the variable is equal to 0, sgn(*) = 0; if the variable is less than 0, sgn(*) = -1. i >ΔT i+1 , indicating that the time interval is decreasing and the transaction frequency may be increasing. At this time, sgn(ΔT i -ΔT i+1 )=1. On the contrary, if ΔTi <ΔT i+1 , indicating that the time interval is increasing and the transaction frequency may be decreasing. At this time, sgn(ΔT i -ΔT i+1 ) = -1. The sign function can capture the directionality of time interval changes, which is important information for evaluating trading behavior trends.

[0059] |ΔT i -ΔT i+1 | α The absolute value difference represents the actual change between adjacent time intervals. By exponentiating α, the degree of influence of the difference on the trend indicator can be adjusted. When α > 1, the impact of the difference is amplified; when 0 < α < 1, the impact of the difference is reduced. This setting allows for flexible adjustment of the trend indicator's sensitivity based on specific application scenarios and needs. In most cases, α is equal to 1, which maximizes the impact of the difference.

[0060] It represents the sum of the differences of all adjacent time intervals and divides it by n-1 (because n time intervals will produce n-1 differences) to obtain the average trend indicator; at the same time, the average processing can smooth out the influence of individual outliers, making the trend indicator more stable and reliable. i -ΔT i+1 The larger it is, the higher the TR1 is, which means the faster the transaction frequency is and the higher the risk is.

[0061] Assume α = 1, and XX Auto Finance Company initiated five auto loan applications in the past year. The time intervals between these five applications are: ΔT1 = 45, ΔT2 = 33, ΔT3 = 21, and ΔT4 = 15. Then calculate the differences: ΔT1 - ΔT2 = 45 - 33 = 12; ΔT2 - ΔT3 = 33 - 21 = 12; and ΔT3 - ΔT4 = 21 - 15 = 6. Substituting these into the equation,

[0062] It should also be noted that in S3, obtaining the second trend indicator based on multiple amount difference data is similar to obtaining the first trend indicator, but there are differences. In S3, obtaining the first trend indicator based on multiple time interval data is expressed as: Among them, TR2 is the second trend indicator, ΔA i is the difference between the i-th adjacent initiated amounts in the amount sequence, ΔA i+1 is the amount difference data between the i+1th adjacent initiated amounts in the amount sequence, n is the number of amount difference data in the amount sequence, and α is the degree of influence.

[0063] The expression for obtaining the second trend indicator is different from that for obtaining the first trend indicator in that the sign function is sgn(ΔAi+1 -ΔA i ), this is because we need to ensure that the higher the TR2, the higher the risk. In the dimension of transaction amount, the higher the risk, the more the transaction amount tends to increase. Therefore, we need to ensure that the variable of the sign function is ΔA. i+1 -ΔA i ; This is different from the trading time dimension. In the trading time dimension, the higher the risk, the more the trading time interval trend will gradually decrease.

[0064] In one embodiment, the first calculation model in obtaining the first parameter based on the first calculation model and the first trend indicator in S2 is expressed as:

[0065] in,

[0066] P1 is the first parameter, TR1 is the first trend indicator, ΔT i is the time interval data between the i-th adjacent initiation time points in the time series, and n is the number of time interval data in the time series.

[0067] In this embodiment, it should be noted that TR1 quantifies the trend of transaction frequency by considering the differences and directionality of adjacent time intervals; a higher TR1 value indicates a larger change in time intervals, which may mean instability or abnormality in transaction behavior, and thus be associated with fraud risk. The average of all time intervals is calculated; this average provides a benchmark for trading frequency, which is used to normalize the first trend indicator. By dividing by the average time interval, the first parameter, P1, reflects the strength of the trend change relative to the average trading frequency. P1 combines the trend indicator and the average time interval to provide a standardized metric for assessing changes in trading behavior over time.

[0068] It should also be noted that in S3, the second calculation model in obtaining the second parameter based on the second calculation model and the second trend indicator is exactly the same as the first calculation model. In S3, the second calculation model in obtaining the second parameter based on the second calculation model and the second trend indicator is expressed as: Among them, P2 is the second parameter, TR2 is the second trend indicator, ΔA i is the amount difference data between the i-th adjacent initiated amounts in the amount sequence, and n is the number of amount difference data in the amount sequence.

[0069] In one embodiment, the correction model in obtaining the adjustment parameter based on the correction model, the first trend indicator, and the second trend indicator in S4 is expressed as:

[0070] in,

[0071] β is the adjustment parameter, t is the standard value of the parameter, TR1 is the first trend indicator, and TR2 is the second trend indicator.

[0072] In this embodiment, it should be noted that TR1 and TR2 represent transaction behavior trends in the time and amount dimensions, respectively. By introducing these two indicators, the revised model can take into account the behavioral changes of transaction entities in different dimensions, thereby more comprehensively assessing fraud risks.

[0073] Max(TR1,1) and max(TR2,1) ensure that the trend indicator is at least 1. This is to filter out the first or second trend indicators that are less than 1. If the first or second trend indicators are negative, there is no risk and they do not need to be included in the calculation of the adjustment parameter, which does not affect the threshold determination. If the first or second trend indicators approach 0, the denominator will be too small, causing the adjustment parameter to be abnormally amplified. Setting a minimum value of 1 ensures the stability of the adjustment parameter.

[0074] and The purpose is to convert the trend indicator into an influencing factor on the adjustment parameter. The larger the trend indicator, the smaller its reciprocal. The smaller it is, the smaller the adjustment parameter is; conversely, the smaller the trend indicator is, the larger its reciprocal is, and the larger the adjustment parameter is.

[0075] The structure of t(1+*) is used to amplify the influence of the trend indicator on the adjustment parameter. When the trend indicator is small, the denominator is small, the amplification factor is equal to 2, and the adjustment parameter is equal to 2t; when the trend indicator is large, the denominator is large, the amplification factor approaches 1, and the adjustment parameter is equal to t. Among them, t is used as the parameter standard value to control the reference range of the adjustment parameter. By adjusting the value of t, the sensitivity and range of the adjustment parameter can be flexibly set to adapt to different application scenarios and needs; preferably, t = 0.5.

[0076] Assume that TR1=10 (calculated based on the time interval data), TR2=15 (calculated based on the amount difference data), and the standard parameter value t=0.5.

[0077] Substituting this into the expression of the modified model, we can calculate that β = 0.5*(1 + 0.167) = 0.5835.

[0078] It should also be noted that the initial probability threshold is corrected according to the adjustment parameter and the target probability threshold is obtained. The adjustment parameter can be directly multiplied by the initial probability threshold. For example, the initial probability threshold is 0.5, the calculated adjustment parameter β is 0.5835, and the target probability threshold is 0.5*0.5835=0.29175.

[0079] In one embodiment, the probability model in obtaining the target fraud probability according to the probability model, the first parameter, and the second parameter in S4 is expressed as:

[0080] in,

[0081] P t is the target fraud probability, P1 is the first parameter, and P2 is the second parameter.

[0082] In this embodiment, it should be noted that exp[-(P1+P2)] converts linear input into nonlinear output. In financial fraud detection, the change of transaction behavior is not linear, so the exponential function can more accurately reflect this nonlinear relationship. The entire expression maps the P1 and P2 inputs to the (0,1) interval, which makes the output P t It can be interpreted as a probability. In financial fraud detection, a probability value is needed to indicate the possibility that a transaction has a fraud risk. P1+P2 in the expression represents the addition of the parameters of the time dimension and the amount dimension. This is because fraudulent behavior is often reflected in multiple dimensions at the same time, such as an abnormal increase in transaction frequency and an abnormal fluctuation in transaction amount. By summing, the information of these two dimensions can be comprehensively considered to more comprehensively assess the fraud risk. The negative sign in -(P1+P2) is to adjust the range of the input P1P2 to the appropriate probability output range. When the input is 0, P t The output is 0.5; when P1+P2 is large (indicating a higher risk of fraud), -(P1+P2) is small, P t The output is close to 1; on the contrary, when P1+P2 is small, -(P1+P2) is large, P t The function output is close to 0.

[0083] A deep learning-based automobile finance fraud detection system is also provided. The system is used to implement any of the above-mentioned deep learning-based automobile finance fraud detection methods in real-time. The system includes:

[0084] An acquisition module is used to acquire the transaction subject associated with the automobile financial transaction data, and obtain historical financial transaction information and historical detection time periods based on the transaction subject. Based on deep learning and historical financial transaction information, the module also acquires the number of initiations, initiation time points, and initiation amounts of automobile financial transaction data associated with the transaction subject during the historical detection time period.

[0085] a first calculation module, configured to sort the multiple initiation time points in chronological order and generate a time series, obtain time interval data between adjacent initiation time points based on the time series, obtain a first trend indicator based on the multiple time interval data, and obtain a first parameter based on the first calculation model and the first trend indicator;

[0086] a second calculation module, configured to sort the multiple initiated amounts in chronological order and generate an amount sequence, obtain amount difference data between adjacent initiated amounts based on the amount sequence, obtain a second trend indicator based on the multiple amount difference data, and obtain a second parameter based on the second calculation model and the second trend indicator;

[0087] The fraud detection module is used to obtain an initial probability threshold based on the number of initiations, obtain an adjustment parameter based on a correction model, a first trend indicator, and a second trend indicator, correct the initial probability threshold according to the adjustment parameter and obtain a target probability threshold, obtain a target fraud probability based on the probability model, the first parameter, and the second parameter, and obtain an automobile finance fraud detection result based on the target fraud probability and the target probability threshold.

[0088] In one embodiment, the acquisition module is further used to: identify the identity information of the transaction subject; use the identity information to access the database or data storage system to retrieve and extract historical financial transaction records associated with the transaction subject; and determine the historical detection time period based on the identity information.

[0089] In one embodiment, the fraud detection module is further used to: determine whether the target fraud probability exceeds the target probability threshold; if so, determine that the transaction subject has a fraud risk in the automobile financial transaction data; if not, determine that the transaction subject has no fraud risk in the automobile financial transactions within the historical detection time period, and obtain a risk interval based on the target probability threshold and the target fraud probability.

[0090] In this embodiment, it should be noted that, regarding the above-mentioned deep learning-based automobile financial fraud detection system, the specific manner of performing operations therein has been described in detail in the embodiment of the deep learning-based automobile financial fraud detection method, and will not be elaborated on here.

[0091] Figure 3 1 is a block diagram of an electronic device for an automobile finance fraud detection method based on deep learning according to an exemplary embodiment. Figure 3 As shown, the electronic device 700 may include: a processor 701 , a memory 702 , and may further include one or more of a multimedia component 703 , an I / O interface 704 (input / output interface), and a communication component 705 .

[0092] The processor 701 is used to control the overall operation of the electronic device 700 to complete all or part of the steps in the deep learning-based automobile finance fraud detection method described above. The memory 702 is used to store various types of data to support the operation of the electronic device 700. This data may include, for example, instructions for any application or method operating on the electronic device 700, as well as application-related data, such as contact information, sent and received messages, images, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 702 or transmitted via the communication component 705. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 704 provides an interface between the processor 701 and other interface modules, which may be a keyboard, a mouse, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC or other 5G, etc., or a combination of one or more thereof, is not limited here. Therefore, the corresponding communication component 705 may include: a Wi-Fi module, a Bluetooth module, an NFC module, etc.

[0093] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the above-mentioned deep learning-based automobile financial fraud detection method.

[0094] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, the program instructions implement the steps of the aforementioned deep learning-based automobile finance fraud detection method. For example, the computer-readable storage medium may be the aforementioned memory 702 including the program instructions. The program instructions may be executed by the processor 701 of the electronic device 700 to implement the aforementioned deep learning-based automobile finance fraud detection method.

[0095] In another exemplary embodiment, a computer program product is also provided, which includes a computer program that can be executed by a programmable device, and has a code portion for executing the above-mentioned deep learning-based automobile financial fraud detection method when executed by the programmable device.

[0096] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the scope of protection of the present disclosure.

[0097] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0098] In addition, the various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A method for detecting automobile finance fraud based on deep learning, characterized in that: include: Obtain the transaction subject associated with the auto finance transaction data, and obtain historical financial transaction information and historical detection time periods based on the transaction subject. Furthermore, based on deep learning and historical financial transaction information, obtain the number of initiations, initiation time points, and initiation amounts of auto finance transaction data associated with the transaction subject within the historical detection time period. sorting the multiple initiation time points in chronological order and generating a time series, obtaining time interval data between adjacent initiation time points according to the time series, obtaining a first trend indicator according to the multiple time interval data, and obtaining a first parameter based on the first calculation model and the first trend indicator; Sorting the multiple initiated amounts in time series order to generate an amount sequence, obtaining amount difference data between adjacent initiated amounts based on the amount sequence, obtaining a second trend indicator based on the multiple amount difference data, and obtaining a second parameter based on the second calculation model and the second trend indicator; Obtaining an initial probability threshold based on the number of initiations, obtaining an adjustment parameter based on the correction model, the first trend indicator, and the second trend indicator, correcting the initial probability threshold based on the adjustment parameter and obtaining a target probability threshold, obtaining a target fraud probability based on the probability model, the first parameter, and the second parameter, and obtaining an auto finance fraud detection result based on the target fraud probability and the target probability threshold; The correction model in obtaining the adjustment parameter based on the correction model, the first trend indicator, and the second trend indicator is expressed as: ;in, To adjust the parameters, is the standard value of the parameter, As the first trend indicator, It is the second trend indicator.

2. The automobile finance fraud detection method based on deep learning according to claim 1 is characterized in that: The acquisition of historical financial transaction information and historical detection time periods based on transaction entities includes: Identify the identity information of the transaction subject; Using identity information to access a database or data storage system to retrieve and extract historical financial transaction records associated with the transaction subject; Determine the historical detection time period based on the identity identification information.

3. The automobile finance fraud detection method based on deep learning according to claim 1 is characterized in that: The obtaining of the automobile finance fraud detection result according to the target fraud probability and the target probability threshold includes: Determine whether the target fraud probability exceeds the target probability threshold; If it exceeds, it is determined that the transaction subject has a fraud risk in the automobile financial transaction data; If it does not exceed, it is determined that there is no fraud risk in the automobile financial transactions of the transaction subject during the historical detection period, and the risk interval is obtained based on the target probability threshold and the target fraud probability.

4. The automobile finance fraud detection method based on deep learning according to claim 1 is characterized in that: The first trend indicator is obtained according to the data of multiple time intervals as follows: ;in, As the first trend indicator, is the time interval data between the i-th adjacent initiation time points in the time series, is the time interval data between the i+1th adjacent initiation time points in the time series, is the number of time interval data in the time series, For the degree of impact.

5. The automobile finance fraud detection method based on deep learning according to claim 1 is characterized in that: The first calculation model in obtaining the first parameter based on the first calculation model and the first trend indicator is expressed as: ;in, is the first parameter, As the first trend indicator, is the time interval data between the i-th adjacent initiation time points in the time series, is the number of time interval data in the time series.

6. The automobile finance fraud detection method based on deep learning according to claim 1, characterized in that: The probability model in obtaining the target fraud probability according to the probability model, the first parameter and the second parameter is expressed as: ;in, is the target fraud probability, is the first parameter, is the second parameter.

7. A deep learning-based automobile finance fraud detection system, characterized in that: The system is used to implement the deep learning-based automobile finance fraud detection method according to any one of claims 1 to 6, and the system includes: An acquisition module is used to acquire the transaction subject associated with the automobile financial transaction data, and obtain historical financial transaction information and historical detection time periods based on the transaction subject. Based on deep learning and historical financial transaction information, the module also acquires the number of initiations, initiation time points, and initiation amounts of automobile financial transaction data associated with the transaction subject during the historical detection time period. a first calculation module, configured to sort the multiple initiation time points in chronological order and generate a time series, obtain time interval data between adjacent initiation time points based on the time series, obtain a first trend indicator based on the multiple time interval data, and obtain a first parameter based on the first calculation model and the first trend indicator; a second calculation module, configured to sort the multiple initiated amounts in chronological order and generate an amount sequence, obtain amount difference data between adjacent initiated amounts based on the amount sequence, obtain a second trend indicator based on the multiple amount difference data, and obtain a second parameter based on the second calculation model and the second trend indicator; The fraud detection module is used to obtain an initial probability threshold based on the number of initiations, obtain an adjustment parameter based on a correction model, a first trend indicator, and a second trend indicator, correct the initial probability threshold according to the adjustment parameter and obtain a target probability threshold, obtain a target fraud probability based on the probability model, the first parameter, and the second parameter, and obtain an automobile finance fraud detection result based on the target fraud probability and the target probability threshold.

8. The deep learning-based automobile finance fraud detection system according to claim 7, characterized in that: The acquisition module is further used for: Identify the identity information of the transaction subject; Using identity information to access a database or data storage system to retrieve and extract historical financial transaction records associated with the transaction subject; Determine the historical detection time period based on the identity identification information.

9. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the deep learning-based automobile finance fraud detection method according to any one of claims 1 to 6.

Images