Financial transaction risk assessment method based on big data

Abstract

The invention relates to the technical field of financial transactions, in particular to a financial transaction risk assessment method based on big data, which comprises the following steps: determining whether to start financial transaction risk assessment based on a transaction numerical value ratio, and calculating a transaction time deviation degree and a transaction place deviation degree of a user to judge a risk type of a transaction; calculating the position concentration degree of the user to determine whether the risk type needs to be further judged, and correcting the risk type based on the order style deviation degree of the user; judging whether the transaction can be passed or not based on the browsing purchase conversion times and the order modification times of the user, and determining whether the judgment of passing the transaction is qualified or not based on the risk characterization value of the user; calculating and generating a transaction passing rate based on the risk type up-regulation times and the total transaction passing times; and secondarily verifying whether the transaction can pass or not based on the transaction passing rate of the user, and sending out an alarm signal under the condition that the transaction cannot pass through the secondary verification. The accuracy of the financial transaction risk assessment method based on the big data is improved.

Description

Technical Field

[0001] This invention relates to the field of financial transaction technology, and in particular to a method for assessing financial transaction risks based on big data. Background Technology

[0002] With the rapid development of fintech, financial transactions are becoming increasingly high-frequency, real-time, and complex. Traditional transaction risk assessment methods face significant challenges. Existing technologies lack correlation analysis between users' dynamic behavioral intentions and long-term risk profiles. Many models only identify and intercept transactions after they are executed or fraudulent activities occur. Traditional risk control systems rarely focus on the deeper factors driving trading decisions, such as users' cognitive biases. Engines based on fixed rules struggle to cope with constantly evolving new fraudulent techniques and complex market abuses. Existing solutions typically manage market risk and operational risk separately, failing to link users' real-time trading requests with their overall portfolio risk profile. Therefore, this field needs a comprehensive financial transaction risk assessment method with self-verification and iterative optimization capabilities.

[0003] Chinese Patent Application Publication No. CN113435770A discloses a blockchain-based transaction risk assessment method and apparatus, relating to the field of blockchain technology. The method includes: extracting the risk type of a target customer based on their historical risk data and storing the extracted risk type on a blockchain network; upon receiving a business transaction request from a target customer at any financial institution, querying the target customer's risk type on the blockchain network and selecting the corresponding risk warning model from a subset of risk warning models for each risk type; inputting the target customer's real-time transaction data into the selected risk warning models and outputting the corresponding risk assessment results; and using a pre-configured smart contract on the blockchain network, determining whether to restrict the target customer's transaction operations based on the target customer's risk type and the corresponding risk assessment results. This invention can obtain more accurate risk assessment results and effectively prevent risky business transactions.

[0004] The existing technology also has the following problems: the existing technology does not take into account the auxiliary verification of risk type assessment results, timely correction of inaccurate risk type assessment results, and exclusion of dangerous users based on the pass rate of each user's transaction. Summary of the Invention

[0005] To address this, the present invention provides a big data-based financial transaction risk assessment method to overcome the problems in existing technologies that do not consider auxiliary verification of risk type assessment results, timely correction of inaccurate risk type assessment results, and exclusion of dangerous users based on the success rate of each user's transaction.

[0006] To achieve the above objectives, this invention provides a financial transaction risk assessment method based on big data, comprising: Based on the transaction value ratio, determine whether to enable financial transaction risk assessment, calculate the user's transaction time deviation and transaction location deviation to generate deviation characterization value to determine the risk type of the transaction; Calculate the user's position concentration to determine whether the risk type needs further judgment. If the risk type needs further judgment, correct the risk type based on the user's order style deviation. If the risk type is determined to be low risk, a transaction feature value is generated based on the number of times the user browses and purchases and the number of times the order is modified to determine whether the transaction can be approved. The user's risk characterization value is used to determine whether the judgment of passing the transaction is qualified. Under the condition that the judgment of this transaction is qualified, the transaction pass rate is calculated based on the number of times the risk type is upgraded and the total number of transactions passed; If the transaction is deemed unqualified, a second verification is performed based on the user's transaction pass rate. If the second verification fails, an alarm signal is issued.

[0007] Furthermore, the determination of whether to initiate a financial transaction risk assessment based on the transaction value ratio, wherein, If the transaction value ratio is greater than or equal to the preset transaction value ratio, then a financial transaction risk assessment is initiated.

[0008] Furthermore, the risk type of a transaction is determined based on the deviation characterization value, whereby... If the deviation characterization value is less than the preset deviation characterization value, the risk type of the transaction is determined to be low risk; If the deviation characterization value is greater than or equal to the preset deviation characterization value, the risk type of the transaction is determined to be high risk.

[0009] Furthermore, based on the user's position concentration, it is determined whether the risk type requires further assessment, wherein... If the position concentration is less than the first preset position concentration, and the risk type of the transaction is determined to be low, then further determination of the risk type is required. If the risk type of the transaction is determined to be low risk, and the position concentration is greater than or equal to the first preset position concentration, then the risk type does not need to be further determined. If the risk type of a transaction is determined to be high risk, and the position concentration is less than or equal to the second preset position concentration, then no further determination is needed to determine the risk type. If the position concentration is greater than the second preset position concentration when the risk type of the transaction is determined to be high risk, then the risk type needs to be further determined. The first preset position concentration is less than the second preset position concentration.

[0010] Furthermore, the risk type is corrected based on the user's order style deviation, wherein, If the risk type is determined to be low risk, and the order style deviation is greater than or equal to the preset order style deviation, then the risk type is determined to be high risk. If the risk type is determined to be high risk, and the order style deviation is less than the preset order style deviation, then the risk type is determined to be low risk.

[0011] Furthermore, the process of generating transaction feature values ​​based on the number of user browsing and purchase conversions and the number of order modifications includes, The conversion ratio is defined as the ratio of the number of conversions to the historical average number of conversions. The ratio of the number of modifications to the historical average number of modifications is defined as the modification ratio. The weighted sum of the conversion rate and the modification rate is determined as the transaction characteristic value.

[0012] Furthermore, the success of the transaction is determined based on transaction characteristic values, among which... If the transaction feature value is less than or equal to the preset transaction feature value, the transaction is deemed to have passed. If the transaction feature value is greater than the preset transaction feature value, the transaction is determined to fail.

[0013] Furthermore, the weighted sum of a user's dependence on a single channel and their dependence on passive push notifications is determined as a risk characterization value.

[0014] Furthermore, the eligibility for approval of this transaction is determined based on the risk characterization value, wherein, If the risk characterization value is less than or equal to the preset risk characterization value, the transaction is deemed to be qualified. If the risk characterization value is greater than the preset risk characterization value, then the transaction is deemed unqualified.

[0015] Furthermore, the secondary verification based on the user's transaction pass rate determines whether the transaction can proceed. If the transaction pass rate is greater than the preset transaction pass rate, then the transaction cannot be approved.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: by determining whether to activate financial transaction risk assessment based on transaction value ratio, this invention saves valuable computing resources and processing time. By calculating the user's transaction time deviation and transaction location deviation, a deviation characterization value is generated to determine the risk type of the transaction. Each user's transaction time and location habits constitute their unique behavioral baseline. Abnormal deviations in transaction time and location are strong correlation signals of high-risk behavior. Real-time calculation of the deviation from the historical baseline can identify malicious behaviors that attempt to evade detection through compliant amounts earlier and more accurately, improving the ability to discover hidden risks, thereby further improving the accuracy of the big data-based financial transaction risk assessment method.

[0017] Furthermore, this invention determines whether the risk type requires further assessment by calculating the user's position concentration. Anomalies in trading time and location may stem from account irregularities or from a legitimate portfolio adjustment by an aggressive investor at an unusual time. Position concentration can identify the latter. If the user already has a large position in the asset, their trading behavior is consistent with the investment logic. Many professional trading strategies naturally generate seemingly abnormal behavioral patterns in terms of time and location, but these users usually have sophisticated position management. Through position concentration verification, the system can recognize that these abnormal behaviors may stem from their professional strategies rather than malicious operations. Under the condition that the risk type requires further assessment, the risk type is corrected based on the user's order style deviation. The order style deviation is used to judge the degree of conformity between the current trading order style and the user's historical habits to correct the trading risk type. By combining static position risk with dynamic behavior, a cross-validation decision-making mechanism is established, thereby further improving the accuracy of the big data-based financial trading risk assessment method.

[0018] Furthermore, this invention, under the condition of determining a low-risk risk type, analyzes the number of times a user browses and purchases to convert and the number of times an order is modified to generate transaction characteristic values ​​to determine whether a user can proceed with the transaction. An extremely low number of browsing and purchasing conversions (i.e., buying immediately after viewing) may represent two risks: first, account theft, where fraudsters have a clear target and place orders directly; second, irrational impulsive transactions by the user. An abnormally high number of order modifications, on the other hand, may indicate market testing, deceptive transactions, or adjustments to attack parameters. By analyzing the combination of these two characteristics, the system can more precisely interpret the intentions behind abnormal transactions. For example, low browsing conversion and high order modification are more likely to indicate market abuse or fraud, while low browsing conversion and no modifications may be more likely to indicate account theft or extreme impulsiveness. This makes risk interception more targeted, bypassing a single user channel. The system uses dependence on single-channel and passive push information to determine the deviation of a user's information acquisition channels and thus assess the validity of a transaction. For users with suspicious transaction characteristics, if both single-channel dependence and passive push dependence do not deviate significantly from historical records, the system can assume there may be a special reason for their transaction and, due to their high long-term credibility, grant them a second verification opportunity. However, users with suspicious transaction characteristics and significant deviations in information behavior are identified as unqualified users, and their transactions are resolutely blocked. This gives the system dynamic credibility assessment capabilities. Furthermore, by analyzing users' single-channel and passive push dependence, the system can identify potential victims exposed to high-risk information environments in advance, thereby further improving the accuracy of big data-based financial transaction risk assessment methods.

[0019] Furthermore, this invention, under the condition that the transaction is deemed unqualified, will query the user's transaction pass rate. If the user's transaction pass rate is high, it means that the failure of this transaction is likely a misjudgment, and the system can change the result to qualified, reducing resource consumption. If the user's transaction pass rate is low, it means that the user is a high-risk individual, and this alarm is highly credible. The system should concentrate resources and immediately issue an alarm signal. The transaction pass rate is not a static value, but is dynamically updated with each transaction of the user. For each qualified transaction result of the user, the system will adjust the user's transaction pass rate by updating the number of times the risk type is increased and the total number of transactions passed to make the next verification result more accurate. The alarm signal issued by the system at the end is the result after being filtered by real-time risk detection and historical credit, thereby further improving the accuracy of the financial transaction risk assessment method based on big data. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the steps of the financial transaction risk assessment method based on big data, as described in an embodiment of the present invention. Figure 2A logic diagram for determining whether to initiate a financial transaction risk assessment in an embodiment of the present invention; Figure 3 This is a logic diagram for determining the risk type of a transaction according to an embodiment of the present invention; Figure 4 This is a logic diagram for determining whether a transaction can proceed according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0024] It is understood that in this embodiment, "user" refers to the investor, subscriber, or buyer in a financial transaction.

[0025] Please see Figure 1 The diagram shows a flowchart illustrating the steps of a big data-based financial transaction risk assessment method according to an embodiment of the present invention. The big data-based financial transaction risk assessment method according to an embodiment of the present invention includes: Step S1: Determine whether to enable financial transaction risk assessment based on the transaction value ratio, calculate the user's transaction time deviation and transaction location deviation to generate a deviation characterization value to determine the risk type of the transaction; Step S2: Calculate the user's position concentration to determine whether the risk type needs further judgment. If the risk type needs further judgment, correct the risk type based on the user's order style deviation. Step S3: Under the condition that the risk type is determined to be low risk, a transaction feature value is generated based on the number of times the user browses and purchases and the number of times the order is modified to determine whether the transaction can be approved. The user's risk characterization value is used to determine whether the judgment of passing the transaction is qualified. Step S4: Under the condition that the judgment of this transaction is qualified, calculate the transaction pass rate based on the number of times the risk type was upgraded and the total number of transactions passed; Step S5: If the transaction is deemed unqualified, a second verification is performed based on the user's transaction pass rate to determine whether the transaction can pass. If the second verification fails, an alarm signal is issued.

[0026] Please see Figure 2 As shown, this is a logic diagram for determining whether to activate financial transaction risk assessment according to an embodiment of the present invention. In step S1, it is determined whether to activate financial transaction risk assessment based on the transaction value ratio. If the transaction value ratio is less than the preset transaction value ratio, the financial transaction risk assessment will not be activated. If the transaction value ratio is greater than or equal to the preset transaction value ratio, then a financial transaction risk assessment is initiated.

[0027] Specifically, the transaction value ratio is the ratio of the current transaction value to the total historical transaction value.

[0028] Understandably, comparing the value of a user's current transaction with the total value of historical transactions can provide a clear picture of the hidden risks in the transaction. Since there are many different business scenarios in financial transactions, in order to balance sensitivity and practicality, and to ensure that it can cover abnormally large transactions without causing frequent evaluations due to an excessively low threshold, thus affecting the user experience, the transaction value ratio is generally preset to a range of 10% to 20%.

[0029] In one specific embodiment, a preset transaction value ratio is set to 15%. If the transaction value ratio is 9%, which is less than the preset transaction value ratio, it is determined that the financial transaction risk assessment will not be activated. If the transaction value ratio is 18%, which is greater than the preset transaction value ratio, then a financial transaction risk assessment will be initiated.

[0030] Please see Figure 3 As shown, this is a logic diagram for determining the risk type of a transaction according to an embodiment of the present invention. The risk type of a transaction is determined based on the deviation characterization value, wherein... If the deviation characterization value is less than the preset deviation characterization value, the risk type of the transaction is determined to be low risk; If the deviation characterization value is greater than or equal to the preset deviation characterization value, the risk type of the transaction is determined to be high risk.

[0031] Specifically, when calculating the user's transaction time deviation, the time period matching degree T1 is quantified; When the trading time is set to T1=0 during regular trading hours, T1=0.3 during less popular trading hours, and T1=0.8 during no trading hours; Understandably, a single trading day can be divided into several time periods, and the number of transactions within each time period can be counted. Time periods with a number of transactions greater than or equal to a preset transaction threshold are designated as regular trading periods, time periods with a number of transactions less than the preset threshold are designated as unpopular trading periods, and time periods with no transactions are designated as no trading periods. The division of time periods and the preset transaction threshold can be based on the market value of the trading entity. For example, if the trading entity is a bicycle with a market value of 500 yuan, the single trading day can be divided into 1-hour intervals, and the preset transaction threshold can be set to 10 times. If the trading entity is a car with a market value of 100,000 yuan, the single trading day can be divided into 2-hour intervals, and the preset transaction threshold can be set to 5 times.

[0032] Quantization frequency deviation T2; Calculate the percentage of this trading session in the total historical trading volume; Frequency deviation T2 = 1 - (the historical percentage of the current time period / the historical highest time period percentage); The transaction time deviation is the weighted sum of the time period matching degree and the frequency deviation degree.

[0033] Understandably, the "historical percentage of this time period" refers to the percentage of transactions that occurred within the same time period as the current transaction out of all transactions made by the user in the past 90 days. The "historical highest time period percentage" refers to the percentage of transactions made during the time period with the most transactions made by the user in the past 90 days. For example, if the total number of transactions made by the user in the past 90 days is 100, and the current transaction occurred in time period A, and the historical number of transactions in time period A is 20, then the historical percentage of this time period = 20 / 100 = 0.2. The historical percentage of time period B is 0.7, and the historical percentage of time period C is 0.1. Therefore, the historical highest time period percentage is 0.7.

[0034] Specifically, the sum of the weighting coefficients for time period matching degree and frequency deviation degree is 1. Since time period matching degree has a greater impact on transaction time deviation degree, the weighting coefficient for time period matching degree is generally taken as 0.6, and the weighting coefficient for frequency deviation degree is 0.4. Only the numerical value is used for calculation when calculating transaction time deviation degree.

[0035] In one specific embodiment, the user's transaction time is during a less popular trading period, so T1=0.3. The historical highest trading period accounts for 0.6, and the current trading period accounts for 0.05, so T2=1-0.05 / 0.6=0.92. Therefore, the transaction time deviation is 0.85.

[0036] Specifically, when calculating the user's transaction location deviation, the location matching degree L1 is quantified; When the location of each trading party is within the top three most frequently used trading locations, L1=0; when the locations of each trading party are in the same city but not within the top three, L1=0.2; when the locations of each trading party are in different cities but in the same province, L1=0.5; when the locations of each trading party are in different provinces, L1=0.8; when the locations of each trading party are in different regions, L1=1.0. Quantify geographic distance deviation L2; Calculate the average distance between the current transaction location and the locations of the three most recent transactions; Set the first distance threshold D0 = 50 km, the second distance threshold D1 = 500 km, and the third distance threshold D2 = 1000 km; Distance deviation ; The transaction location deviation is the weighted sum of the location matching degree and the geographical distance deviation.

[0037] Specifically, the sum of the weighting coefficients for location matching degree and geographical distance deviation degree is 1. Since location matching degree has a greater impact on transaction location deviation degree, the weighting coefficient for time period matching degree is generally taken as 0.7, and the weighting coefficient for geographical distance deviation degree is 0.3. Only the numerical value is used for calculation when calculating transaction location deviation degree.

[0038] In one specific embodiment, the user's transaction location spans multiple provinces, so L1=0.8. The average distance between the current transaction location and the locations of the three most recent transactions is 800 kilometers, so D1=500<800≤D2=1000. L2=(800-500) / (1000-500)+0.5=1.1, which is adjusted to the upper limit of 1.0. Therefore, the transaction location deviation is 0.86.

[0039] Specifically, the weighted sum of the transaction time deviation and the transaction location deviation is determined as the deviation characterization value.

[0040] Specifically, the sum of the weighting coefficients for transaction time deviation and transaction location deviation is 1. Since transaction time deviation and transaction location deviation have the same impact on the risk type of a transaction, the weighting coefficient for transaction time deviation is generally taken as 0.5, and the weighting coefficient for transaction location deviation is also 0.5.

[0041] In one specific embodiment, a preset deviation characterization value is set to 0.65. If the deviation characterization value is 0.62, which is less than the preset deviation characterization value, the risk type of the transaction is determined to be low risk. If the deviation characterization value is 0.68, which is greater than the preset deviation characterization value, the risk type of the transaction is determined to be high risk.

[0042] Understandably, if only the transaction time or the transaction location is different, the overall deviation will not be too large, indicating that the user has only accidentally performed an unconventional operation. However, when both the transaction time and the transaction location are different, resulting in an overall deviation, it indicates that the user is in an abnormal risk situation. Therefore, the default deviation value is generally set to a range of 0.6 to 0.7.

[0043] Specifically, this invention saves valuable computing resources and processing time by determining whether to activate financial transaction risk assessment based on transaction value ratios. It generates deviation characterization values ​​by calculating the user's transaction time deviation and transaction location deviation to determine the risk type of the transaction. Each user's transaction time and location habits constitute their unique behavioral baseline. Abnormal deviations in transaction time and location are strong correlation signals of high-risk behavior. Real-time calculation of the deviation from the historical baseline can identify malicious behaviors that attempt to evade detection through compliant amounts earlier and more accurately, improving the ability to discover hidden risks and thus further improving the accuracy of the big data-based financial transaction risk assessment method.

[0044] Specifically, whether further determination is needed to classify the risk type based on the user's position concentration is as follows: If the risk type of a transaction is determined to be low, and the position concentration is less than the first preset position concentration, then the risk type needs to be further determined. If the risk type of a transaction is determined to be low risk, and the position concentration is greater than or equal to the first preset position concentration, then no further determination of the risk type is required. If the risk type of a transaction is determined to be high risk, and the position concentration is less than or equal to the second preset position concentration, then no further determination of the risk type is required. If the risk type of a transaction is determined to be high risk, and the position concentration is greater than the second preset position concentration, then the risk type needs to be further determined. The first preset position concentration is less than the second preset position concentration.

[0045] Specifically, the weighted sum of single asset concentration, industry concentration, and liquidity concentration is the position concentration.

[0046] Specifically, single asset concentration is the ratio of the largest asset market value to the total asset value, industry concentration is the ratio of the largest industry market value to the total asset value, and liquidity concentration is the ratio of the value of illiquid assets to the total asset value. For those skilled in the art, the statistics or calculation of the largest asset market value, total asset value, largest industry market value, total asset value, value of illiquid assets, and total asset value are existing technologies and will not be elaborated here.

[0047] Understandably, single asset concentration measures the proportion of a user's holdings in the largest single asset, reflecting the risk of over-reliance on a specific asset. Industry concentration measures the concentration risk of a user's holdings at the industry level, while liquidity concentration measures the proportion of assets in a user's holdings that are difficult to liquidate quickly.

[0048] Specifically, the sum of the weighting coefficients for single asset concentration, industry concentration, and liquidity concentration is 1. Since single asset concentration has a greater impact on position concentration, while industry concentration and liquidity concentration have a relatively smaller impact on position concentration, the weighting coefficients for single asset concentration, industry concentration, and liquidity concentration are generally taken as 0.4, 0.3, and 0.3, respectively.

[0049] Understandably, the core characteristics of low-risk transactions are diversified allocation and low risk exposure. If the position concentration is less than the first preset position concentration, it means that the user's funds are diversified across multiple assets and industries, which aligns with the nature of low risk. If the position concentration is greater than the first preset position concentration, it indicates that funds are becoming overly concentrated, contradicting the low-risk assessment. The core characteristics of high-risk transactions are concentrated exposure and large risk exposure. If the position concentration is greater than the second preset position concentration, it means that the user's funds are highly concentrated, aligning with the nature of high risk. If the position concentration is less than the second preset position concentration, it indicates that funds are still diversified, contradicting the high-risk assessment. The position concentration is determined based on the number of different types of trading entities in the user's historical financial transactions. Optionally, in this embodiment, the first preset position concentration ranges from 0.3 to 0.35, and the second preset position concentration ranges from 0.7 to 0.75.

[0050] Specifically, risk types are corrected based on the user's order style deviation, where... If the risk type is determined to be low risk, and the order style deviation is less than the preset order style deviation, then the risk type will not be corrected. If the order style deviation is greater than or equal to the preset order style deviation, and the risk type is determined to be high risk, then the risk type to be corrected is determined to be low risk. If the order style deviation is less than the preset order style deviation when the risk type is determined to be high risk, then the risk type to be corrected is determined to be low risk. If the risk type is determined to be high risk, and the order style deviation is greater than or equal to the preset order style deviation, then the risk type will not be corrected.

[0051] Specifically, the weighted sum of order type deviation and order size deviation is used to calculate the order style deviation degree.

[0052] Specifically, the order type deviation = 1 - the historical frequency of this order type, and the order size deviation = 1 - |this transaction amount - the average historical transaction amount| / the average historical transaction amount.

[0053] It is understandable that the order type refers to the specific method by which a user sets the transaction price and conditions when placing a transaction instruction. Common order types include market orders, limit orders, and stop-loss orders. If the user's transaction is a market order, then the historical frequency of market orders is 20 / 100=0.2, and the order type deviation is 1-0.2=0.8.

[0054] Understandably, order type deviation measures the difference between the type of order a user is currently placing and the type of order they have historically preferred, while order size deviation measures the degree of deviation between the amount of the order a user is currently placing and the amount of the order they have historically preferred.

[0055] Specifically, the sum of the weighting coefficients for order type deviation and order size deviation is 1. Since order size deviation has a greater impact on the overall order style deviation than order type deviation, the weighting coefficient for order type deviation is generally taken as 0.4, and the weighting coefficient for order size deviation is 0.6.

[0056] Understandably, the order style deviation is determined by both the order type deviation and the order size deviation. When one dimension has a large deviation, the impact on the overall order style deviation will not be significant. However, when one dimension has a large deviation, it indicates that the user's transaction style is abnormal compared to usual and should raise a red flag. Therefore, the default value range for the order style deviation is generally 0.6 to 0.65.

[0057] Specifically, this invention determines whether the risk type requires further assessment by calculating the user's position concentration. Anomalies in trading time and location may stem from account irregularities or from a legitimate portfolio adjustment by an aggressive investor at an unusual time. Position concentration can identify the latter. If the user already has a large position in the asset, their trading behavior is consistent with their investment logic. Many professional trading strategies naturally generate seemingly abnormal behavioral patterns in terms of time and location, but these users usually have sophisticated position management. Through position concentration verification, the system can recognize that these abnormal behaviors may stem from their professional strategies rather than malicious operations. When it is determined that the risk type requires further assessment, the risk type is corrected based on the user's order style deviation. The order style deviation is used to determine the degree of conformity between the current trading order style and the user's historical habits to correct the trading risk type. By combining static position risk with dynamic behavior, a cross-validation decision-making mechanism is established, thereby further improving the accuracy of the big data-based financial trading risk assessment method.

[0058] Specifically, the process of generating transaction feature values ​​based on a user's browsing and purchase conversion frequency and order modification frequency includes, The conversion ratio is defined as the ratio of the number of conversions to the historical average number of conversions. The ratio of the number of modifications to the historical average number of modifications is defined as the modification ratio. The weighted sum of the conversion rate and the modification rate is determined as the transaction characteristic value.

[0059] Understandably, the number of browsing-to-purchase conversions refers to the number of times a user goes from browsing a product or asset page to successfully completing a purchase within a specific time period, such as one day. It reflects the speed of the user's purchase decision and the clarity of their purchase intention. The number of order modifications refers to the total number of times a user modifies the order details, such as quantity, price, and type, after submitting the order and before final payment in a single transaction process. It reflects the user's degree of hesitation and uncertainty.

[0060] Understandably, the average number of historical conversions and the average number of historical modifications are automatically collected and stored by the platform during user usage and are considered core data assets of the platform. For example, if a user's daily conversions over the past 5 days were 2, 3, 1, 0, and 4, then their current average number of historical conversions = (2+3+1+0+4) / 5 = 2. If a user has completed 3 transactions in the past, with each transaction modified 1, 5, and 3 times respectively, then their current average number of historical modifications = (1+5+3) / 3 = 3.

[0061] Specifically, the sum of the weighting coefficients for the conversion rate and the modification rate is 1. Since frequent modifications to order information may be due to fraud or falsification, the modification rate is more sensitive to transaction risks than the conversion rate. Therefore, the weighting coefficient for the conversion rate is generally 0.3, and the weighting coefficient for the modification rate is 0.7.

[0062] Specifically, the determination of whether a transaction can proceed is based on transaction characteristic values, among which... If the transaction characteristic value is less than or equal to the preset transaction characteristic value, the transaction is deemed to have passed. If the transaction characteristic value is greater than the preset transaction characteristic value, the transaction is deemed unsuccessful.

[0063] Understandably, the browsing-to-purchase conversion rate measures the rationality of a user's purchasing decision, while the order modification rate measures the stability of a user's order information. Together, they reflect the authenticity and compliance of the transaction.

[0064] In one specific embodiment, a preset transaction feature value of 0.98 is set. If the transaction feature value of 0.95 is less than or equal to the preset transaction feature value, the transaction is determined to be successful. If the transaction feature value is 0.98, which is greater than the preset transaction feature value, then the transaction is deemed unsuccessful.

[0065] It is understandable that user conversion frequency will increase significantly during promotional periods, but high-risk scenarios have extremely high requirements for conversion rate and order stability, and abnormal behavior needs to be strictly intercepted. Therefore, the default value range for transaction characteristics is generally 0.95 to 1.15.

[0066] Specifically, the weighted sum of a user's dependence on a single channel and their dependence on passive push notifications is determined as the risk characterization value.

[0067] Specifically, the eligibility for approval of this transaction is determined based on the risk characterization value. If the risk characterization value is less than or equal to the preset risk characterization value, the transaction is deemed to have passed the assessment. If the risk characterization value is greater than the preset risk characterization value, the transaction will be deemed unqualified.

[0068] Understandably, reliance on a single channel measures the risk of information source concentration. In financial transactions, a high reliance on a single channel may prevent users from obtaining comprehensive and balanced market information, making them susceptible to being misled or manipulated by a single source. Reliance on passive push measures the initiative of users in obtaining information. Active search usually represents having clear goals and rational thinking. For users who rely on algorithm push, their information environment is passively shaped, and the platform algorithm will prioritize pushing content that attracts attention and evokes emotions.

[0069] Specifically, single-channel dependence is the ratio of the number of information items from the user's most frequently used channel to the total number of information items obtained, while passive push dependence is the ratio of the number of information items passively obtained to the total number of information items obtained. Therefore, the higher the single-channel dependence and the higher the passive push dependence, the greater the deviation between the user's information acquisition channels on the financial trading platform and their information acquisition behavior within those channels.

[0070] Specifically, the sum of the weighting coefficients of single-channel dependence and passive push dependence is 1. Since single-channel dependence and passive push dependence have the same impact on the overall transaction risk index, the weighting coefficient of single-channel dependence is generally taken as 0.5, and the weighting coefficient of passive push dependence is also 0.5.

[0071] In one specific embodiment, a preset risk characterization value is set to 0.8. If the risk characterization value is 0.73, which is less than the preset risk characterization value, then the transaction is deemed to be qualified. If the risk characterization value is 0.87, which is greater than the preset risk characterization value, then the transaction is deemed unqualified.

[0072] It is understandable that users with multiple information acquisition channels and proactive information acquisition indicate low perceived risk, while users with a single information acquisition channel and passive information acquisition indicate high perceived risk. For users with only mild bias, it may indicate that the user simply likes to watch push notifications but the channels are relatively reliable. However, blocking only accounts with extreme risks will result in weak risk control. Therefore, the preset risk characterization value range is 0.75 to 0.9.

[0073] Specifically, this invention, under the condition of determining a low-risk risk type, analyzes the number of times a user browses and purchases to convert and the number of times an order is modified to generate transaction characteristic values ​​to determine whether a user can proceed with the transaction. An extremely low number of browsing and purchasing conversions (i.e., buying immediately after viewing) may represent two risks: first, account theft, where fraudsters have a clear target and place orders directly; second, irrational impulsive transactions by the user. An abnormally high number of order modifications, on the other hand, may indicate market testing, deceptive transactions, or adjustments to attack parameters. By analyzing the combination of these two characteristics, the system can more precisely interpret the intentions behind abnormal transactions. For example, low browsing conversion and high order modification are more likely to indicate market abuse or fraud, while low browsing conversion and no modifications may be more likely to indicate account theft or extreme impulsiveness. This makes risk interception more targeted, bypassing a single user channel. The system uses dependence on single-channel and passive push information to determine the deviation of a user's information acquisition channels and thus assess the validity of a transaction. For users with suspicious transaction characteristics, if both single-channel dependence and passive push dependence do not deviate significantly from historical records, the system can assume there may be a special reason for their transaction and, due to their high long-term credibility, grant them a second verification opportunity. However, users with suspicious transaction characteristics and significant deviations in information behavior are identified as unqualified users, and their transactions are resolutely blocked. This gives the system dynamic credibility assessment capabilities. Furthermore, by analyzing users' single-channel and passive push dependence, the system can identify potential victims exposed to high-risk information environments in advance, thereby further improving the accuracy of big data-based financial transaction risk assessment methods.

[0074] Please see Figure 4 As shown, this is a logic diagram for determining whether a transaction can proceed according to an embodiment of the present invention. The transaction success rate is used for secondary verification to determine whether the transaction can proceed. If the transaction success rate is less than or equal to the preset transaction success rate, the transaction will be deemed unsuccessful. If the transaction success rate is greater than the preset success rate, the transaction will be deemed unsuccessful.

[0075] Specifically, the transaction approval rate is the ratio of the number of times the risk type was upgraded to the total number of transactions approved.

[0076] Understandably, after a transaction is deemed unqualified, the system proceeds to a second verification. At this point, the system deals with high-risk, suspicious transactions. If a user's historical transaction pass rate is high, it indicates that they generally tend to make normal transactions. However, this transaction was deemed unqualified by the previous rules, which is an abnormal event, so it can be allowed to proceed. If a user's historical pass rate is low, it indicates that they generally tend to make suspicious transactions, and an alarm signal needs to be issued. Therefore, the default transaction pass rate is generally set between 95% and 98%.

[0077] Specifically, this invention, when determining that a transaction is unqualified, queries the user's transaction pass rate. If the user's pass rate is high, it means the transaction failure was likely a misjudgment, and the system can change the result to qualified, reducing resource consumption. If the user's pass rate is low, it means the user is inherently high-risk, and the alert is highly credible; the system should concentrate resources and immediately issue an alarm signal. The transaction pass rate is not a static value but dynamically updated with each of the user's transactions. For each qualified transaction, the system adjusts the user's pass rate by updating the number of risk type increases and the total number of successful transactions to make the next verification result more accurate. The final alarm signal issued by the system is the result of dual filtering by real-time risk detection and historical creditworthiness, thereby further improving the accuracy of the big data-based financial transaction risk assessment method.

[0078] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for assessing financial transaction risk based on big data, characterized in that, include: Based on the transaction value ratio, determine whether to enable financial transaction risk assessment, calculate the user's transaction time deviation and transaction location deviation to generate deviation characterization value to determine the risk type of the transaction; Calculate the user's position concentration to determine whether the risk type needs further judgment. If the risk type needs further judgment, correct the risk type based on the user's order style deviation. If the risk type is determined to be low risk, a transaction feature value is generated based on the number of times the user browses and purchases and the number of times the order is modified to determine whether the transaction can be approved. The user's risk characterization value is used to determine whether the judgment of passing the transaction is qualified. Under the condition that the judgment of this transaction is qualified, the transaction pass rate is calculated based on the number of times the risk type is upgraded and the total number of transactions passed; If the transaction is deemed unqualified, a second verification is performed based on the user's transaction pass rate. If the second verification fails, an alarm signal is issued.

2. The financial transaction risk assessment method based on big data according to claim 1, characterized in that, The determination of whether to initiate a financial transaction risk assessment based on the transaction value ratio, wherein... If the transaction value ratio is greater than or equal to the preset transaction value ratio, then a financial transaction risk assessment is initiated.

3. The financial transaction risk assessment method based on big data according to claim 2, characterized in that, The risk type of a transaction is determined based on the deviation characterization value, among which, If the deviation characterization value is less than the preset deviation characterization value, the risk type of the transaction is determined to be low risk; If the deviation characterization value is greater than or equal to the preset deviation characterization value, the risk type of the transaction is determined to be high risk.

4. The financial transaction risk assessment method based on big data according to claim 3, characterized in that, Whether the risk type requires further determination is determined based on the user's position concentration. If the position concentration is less than the first preset position concentration, and the risk type of the transaction is determined to be low, then further determination of the risk type is required. If the risk type of the transaction is determined to be low risk, and the position concentration is greater than or equal to the first preset position concentration, then the risk type does not need to be further determined. If the risk type of a transaction is determined to be high risk, and the position concentration is less than or equal to the second preset position concentration, then no further determination is needed to determine the risk type. If the position concentration is greater than the second preset position concentration when the risk type of the transaction is determined to be high risk, then the risk type needs to be further determined. The first preset position concentration is less than the second preset position concentration.

5. The financial transaction risk assessment method based on big data according to claim 4, characterized in that, The risk type is corrected based on the user's order style deviation, wherein... If the risk type is determined to be low risk, and the order style deviation is greater than or equal to the preset order style deviation, then the risk type is determined to be high risk. If the risk type is determined to be high risk, and the order style deviation is less than the preset order style deviation, then the risk type is determined to be low risk.

6. The financial transaction risk assessment method based on big data according to claim 5, characterized in that, The process of generating transaction feature values ​​based on the number of user browsing and purchase conversions and the number of order modifications includes: The conversion ratio is defined as the ratio of the number of conversions to the historical average number of conversions. The ratio of the number of modifications to the historical average number of modifications is defined as the modification ratio. The weighted sum of the conversion rate and the modification rate is determined as the transaction characteristic value.

7. The financial transaction risk assessment method based on big data according to claim 6, characterized in that, The determination of whether a transaction can proceed is based on transaction characteristic values, among which... If the transaction feature value is less than or equal to the preset transaction feature value, the transaction is deemed to have passed. If the transaction feature value is greater than the preset transaction feature value, the transaction is determined to fail.

8. The financial transaction risk assessment method based on big data according to claim 7, characterized in that, The weighted sum of a user's dependence on a single channel and their dependence on passive push notifications is determined as the risk characterization value.

9. The financial transaction risk assessment method based on big data according to claim 8, characterized in that, The eligibility of this transaction is determined based on the risk characterization value. If the risk characterization value is less than or equal to the preset risk characterization value, the transaction is deemed to be qualified. If the risk characterization value is greater than the preset risk characterization value, then the transaction is deemed unqualified.

10. The financial transaction risk assessment method based on big data according to claim 9, characterized in that, The secondary verification based on the user's transaction success rate determines whether the transaction can proceed. If the transaction pass rate is greater than the preset transaction pass rate, then the transaction cannot be approved.

Images