A lottery ticket redemption method and system based on time series analysis

Abstract

The application discloses a lottery ticket redemption method and system based on time series analysis, and relates to the technical field of automatic redemption. The method comprises the following steps: extracting the timestamp, location record and redemption frequency of the lottery ticket redemption request, converting the extracted data into a data format and standardizing, and generating a standardized data set. In the application, abnormal data points are further screened out through deviation analysis of individual redemption frequency and overall average redemption frequency, potential risk transactions are identified, the data-driven decision tree training process can determine effective classification thresholds through cross-validation, normal and abnormal behaviors are distinguished, the application of the classification model not only improves the risk assessment efficiency of the current redemption request, but also identifies high-risk transactions in real time through comparison with the preset risk threshold, the introduction of the automatic notification function ensures that the risk transaction information is timely reported to the supervision department and the sales point, and the safety and fairness of the lottery ticket redemption process are strengthened through iterative review and transaction prevention.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automated prize redemption, and in particular to a lottery prize redemption method and system based on time series analysis. BACKGROUND

[0002] The technical field of automated prize redemption involves the use of computer programs and machine learning algorithms to verify and redeem lottery winnings. The core goal is to improve the efficiency and security of lottery prize redemption, reducing the risk of human error and irregular behavior. The automated system scans the barcode or QR code of the lottery ticket, automatically reads the winning information, and compares it with the database of the lottery management center to confirm the authenticity and winning status of the lottery ticket. The automated prize redemption system can also update the winning data in real time, providing accurate winning statistical information for lottery issuers.

[0003] Existing automated prize redemption technology mainly relies on standard applications of machine learning algorithms and database management systems, such as barcode scanning and automatic reading of winning information, limiting the identification of deep-seated abnormal patterns of redemption behavior. Although it can improve the efficiency of redemption and reduce human error, it still lacks the ability to handle more complex irregular behaviors such as location and time anomalies, frequency anomalies, etc. Traditional methods are not flexible enough when updating winning data, affecting the real-time analysis and monitoring of winning data by lottery issuers. These shortcomings result in a slow response from regulatory authorities and sales points when dealing with potential risk transactions, increasing the irregular risks faced by the lottery industry. SUMMARY

[0004] The purpose of the present application is to solve the shortcomings of the prior art and propose a lottery prize redemption method and system based on time series analysis.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solution: a lottery prize redemption method based on time series analysis, comprising the following steps:

[0006] Extract the timestamp, redemption location record, and redemption frequency of the lottery redemption request, convert the extracted data to a data format and standardize it to generate a standardized data set;

[0007] Perform time series analysis on the timestamp of the standardized data set to identify abnormal deviations from the typical redemption time, calculate the distance between the redemption location and the known redemption location based on the abnormal deviations of the redemption time, and obtain the abnormal index analysis result; the typical redemption time is a preset value of the redemption time determined according to normal redemption behavior;

[0008] According to the abnormal index analysis result, calculate the deviation of the individual redemption frequency from the overall average redemption frequency, identify data points exceeding twice the standard deviation as risk transactions, and obtain the frequency deviation statistical result;

[0009] The frequency deviation statistical result is combined with the abnormal index analysis result, decision tree input data is created through data integration, a decision tree training process is executed, a classification threshold is determined through cross validation, potential abnormal and normal prize redemption behaviors are distinguished, and a classification model is constructed;

[0010] The classification model is used to analyze data of a current prize redemption request, calculate an abnormal probability of each transaction, compare the abnormal probability with a preset risk threshold, identify and mark transactions exceeding the threshold, and generate an abnormal probability score table;

[0011] According to the abnormal probability score table, transactions marked as risks are screened, an automatic notification function is used to send an alarm to associated regulatory departments and sales points, measures such as iterative review and transaction blocking are taken, and a prize redemption monitoring result is obtained.

[0012] A lottery prize redemption system based on time series analysis, which is used to execute the above lottery prize redemption method based on time series analysis, the system comprises:

[0013] A data extraction module is configured to extract timestamps, prize redemption location records and prize redemption frequencies of lottery prize redemption requests, convert the extracted data into a data format and standardize the data, and generate a standardized data set;

[0014] A time series analysis module is configured to perform time series analysis on the timestamps of the standardized data set, identify abnormal deviations from typical prize redemption times, calculate distances between prize redemption locations and known prize redemption locations according to the abnormal deviations of the prize redemption times, and obtain abnormal index analysis results; the typical prize redemption time is a prize redemption time preset value determined according to normal prize redemption behavior;

[0015] A frequency analysis module is configured to calculate deviations of individual prize redemption frequencies from overall average prize redemption frequencies according to the abnormal index analysis results, identify data points exceeding twice the standard deviation as risk transactions, and obtain frequency deviation statistical results;

[0016] A decision tree modeling module is configured to use the frequency deviation statistical results in combination with the abnormal index analysis results, create decision tree input data through data integration, execute a decision tree training process, determine a classification threshold through cross validation, distinguish potential abnormal and normal prize redemption behaviors, and construct a classification model;

[0017] An abnormal monitoring module is configured to use the classification model to analyze data of a current prize redemption request, calculate an abnormal probability of each transaction, compare the abnormal probability with a preset risk threshold, identify and mark transactions exceeding the threshold, and generate an abnormal probability score table;

[0018] A risk notification module is configured to screen transactions marked as risky according to the abnormal probability scoring table, send an alert to the associated regulatory department and sales point through an automatic notification function, take measures of iterative review and blocking transactions, and obtain a prize redemption monitoring result.

[0019] Compared with the prior art, the application has the advantages and positive effects that:

[0020] In the application, the ability to identify potential risky transactions is optimized by combining time stamp, location record and redemption frequency analysis, and standardizing data. Abnormal deviations compared with regular redemption times are identified through time series analysis, and distances from known redemption locations are calculated to enhance the positioning accuracy of abnormal redemption activities. Deviation analysis of individual redemption frequency from the overall average redemption frequency further filters out abnormal data points and identifies potential risky transactions. The data-driven decision tree training process can determine effective classification thresholds through cross-validation to distinguish between normal and abnormal behavior. The application of the classification model not only improves the risk assessment efficiency of current redemption requests, but also identifies high-risk transactions in real time by comparing with preset risk thresholds. The introduction of the automatic notification function ensures that risk transaction information is timely reported to the regulatory department and sales point, and the security and fairness of the lottery redemption process are enhanced through iterative review and blocking transactions. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The figure is a workflow diagram of the application;

[0022] Figure 2 The figure is a S1 refinement flowchart of the application;

[0023] Figure 3 The figure is a S2 refinement flowchart of the application;

[0024] Figure 4 The figure is a S3 refinement flowchart of the application;

[0025] Figure 5 The figure is a S4 refinement flowchart of the application;

[0026] Figure 6 The figure is a S5 refinement flowchart of the application;

[0027] Figure 7 The figure is a S6 refinement flowchart of the application;

[0028] Figure 8 The figure is a system flowchart of the application. DETAILED DESCRIPTION

[0029] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0030] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0031] Please refer to Figure 1 The present application provides a technical solution, a lottery ticket redemption method based on time series analysis, comprising the following steps:

[0032] S1: Extracting the timestamp, redemption location record and redemption frequency of the lottery ticket redemption request, converting the extracted data into data format and standardizing, and generating a standardized data set;

[0033] S2: Time series analysis of the timestamp of the standardized data set, identifying the abnormal deviation from the typical redemption time, calculating the distance between the redemption location and the known redemption location according to the abnormal deviation of the redemption time, and obtaining the abnormal index analysis result;

[0034] S3: According to the abnormal index analysis result, calculate the deviation of individual redemption frequency in the lottery ticket redemption frequency data from the overall average redemption frequency, identify the data points exceeding two standard deviations as risk transactions, and obtain the frequency deviation statistical result;

[0035] S4: Using the frequency deviation statistical result, combining the abnormal index analysis result, creating decision tree input data through data integration, executing the decision tree training process, determining the classification threshold through cross-validation, distinguishing between potential abnormal and normal redemption behaviors, and building a classification model;

[0036] S5: Using the classification model, analyzing the data of the current redemption request, calculating the abnormal probability of each transaction, and comparing the abnormal probability with the preset risk threshold, identifying and marking the transactions exceeding the threshold, and generating an abnormal probability score table;

[0037] S6: According to the abnormal probability scoring table, screen the transactions marked as risk, send the alarm to the relevant regulatory departments and sales points through the automatic notification function, take measures of iterative review and blocking transactions, and get the prize monitoring results.

[0038] The standardized data set includes the timestamp of lottery prize redemption, the latitude and longitude identification of the prize redemption location, and the prize redemption frequency value of each prize redemptions, the abnormal index analysis result includes the deviation value of each prize redemption timestamp and the average prize redemption time, the spatial distance of the prize redemption location and the common location, and the abnormal score, the frequency deviation statistical result includes the comparison value of the individual prize redemption frequency and the average prize redemption frequency, the standard deviation calculation result and the identified risk transaction mark, the classification model includes the decision tree model structure trained based on the input data, the classification threshold and the consistency index of model verification, the abnormal probability scoring table includes the transaction number, the calculated abnormal probability and the mark of whether each transaction exceeds the risk threshold, and the prize monitoring result includes the list of marked risk transactions, the content of issued automatic alarm and the taken review and blocking measures.

[0039] Please refer to Figure 2 , extract the timestamp of lottery prize redemption request, prize redemption location record and prize redemption frequency, convert the extracted data into data format and standardize, and the steps of generating the standardized data set are as follows:

[0040] S101: Load and read the lottery prize redemption picture file, apply image recognition technology to identify and record the timestamp and location information in the picture one by one, and the execution process of getting the initialization data record is as follows:

[0041] Load and read the lottery prize redemption picture file, identify and record the timestamp and location information in the picture one by one by applying image recognition technology, the process involves image preprocessing such as image denoising, color adjustment and contrast enhancement, feature extraction is performed on the processed image through neural network models suitable for text recognition such as convolutional neural network (CNN) or recurrent neural network (RNN), each pixel in the image will be analyzed to determine whether it belongs to the part of the timestamp and location information that needs to be extracted, through the training of these models, the system can identify different formats and styles of text, ensuring accurate recording of timestamp and location information, and getting the initialization data record.

[0042] S102: Based on the initialization data record, traverse and check each record, eliminate records with incorrect format and incomplete data, and for the remaining data, count the number of prize redemptions at each location, and the execution process of generating the location prize redemption frequency statistical table is as follows:

[0043] Count the number of prize redemptions at each location, and calculate the prize redemption frequency of each location according to the formula . In the formula, The number of prize redemptions for a location, The location identifier, The location identifier in the record, The indicator function takes the value 1 when and 0 otherwise, The total number of records. Detailed formula and formula calculation derivation process: set a record list containing location identifiers as follows: [A, A, B, C, A, B], calculate the prize redemption number of location A. According to the prize redemption number of location formula, set , traverse the entire list, and whenever , the value of the indicator function is 1, otherwise 0. Therefore, the prize redemption number of location A is calculated as follows: . This shows that the prize redemption number of location A is 3, which indicates that the prize redemption activity of location A is relatively frequent in this sample.

[0044] S103: Format unification and scale adjustment of data in the prize redemption frequency statistics table of the location, check each data according to the unified standard format, and generate the execution process of the standardized data set as follows;

[0045] Format unification and scale adjustment of data in the prize redemption frequency statistics table of the location, the process includes converting the prize redemption number of each location to a standard format, such as converting all records to uppercase or lowercase to eliminate data errors caused by inconsistent formats, scale adjustment involves data normalization, such as converting the prize redemption number to a value between 0 and 1 through the max-min normalization method, which helps improve the calculation efficiency and accuracy in the subsequent data processing and analysis stage, for higher-level data analysis such as trend prediction and pattern recognition to identify which locations have abnormal prize redemption activities or need further review, and generate a standardized data set.

[0046] Please refer to Figure 3 , time series analysis of the time stamp of the standardized data set, identify the abnormal deviation from the typical prize redemption time, calculate the distance between the prize redemption location and the known prize redemption location according to the abnormal deviation of the prize redemption time, and get the steps of the abnormal index analysis result as follows:

[0047] S201: Extract time stamp data from the standardized data set, sequence each time stamp, convert it to continuous time series data, identify the deviation from the typical prize redemption time through statistical analysis of the time series, and generate the execution process of the time deviation analysis result as follows;

[0048] Sequence each time stamp, convert it to continuous time series data, calculate the sequenced time value according to the formula . In the formula, represents the original time stamp, and These represent the minimum and maximum values ​​in the timestamp data, respectively. This is the serialized time value. The typical prize redemption time is a preset value determined based on normal prize redemption behavior. Detailed formula explanation and calculation derivation: Set the timestamp data to: [2023-01-01 10:00, 2023-01-02 12:00, 2023-01-01 11:00], convert to UNIX timestamps. (2023-01-01 10:00) (2023-01-02 12:00). Calculate the serialization time value for 2023-01-01 11:00 by applying the serialization formula:

[0049] This indicates that the relative distance between this timestamp and the smallest timestamp is 0.0833, a value that is helpful for subsequent time series analysis.

[0050] S202: Based on the time deviation analysis results, calculate the degree of deviation between each data point in the time series and the typical prize redemption time, mark the abnormalities according to the magnitude of the deviation, and mark the time points where the deviation exceeds the set threshold. The execution process for obtaining the abnormal time marking results is as follows.

[0051] Based on the time deviation analysis results, the process includes using statistical methods to calculate the degree of deviation between each data point in the time series and the typical prize redemption time, and marking anomalies according to the size of the deviation. By setting a threshold, for example, a deviation value exceeding 24 hours is considered an anomaly. The threshold is determined based on the statistical analysis of historical data. The anomaly marking method can be a simple conditional judgment, such as marking an anomaly if the deviation is greater than the set 24 hours. Although this method is simple, it is very effective in practical applications, and the abnormal time marking results are obtained.

[0052] S203: The execution flow for determining the prize redemption location corresponding to the abnormal time stamp result as the abnormal lottery prize redemption location, calculating the distance between the abnormal lottery prize redemption location and the known prize redemption location, summarizing the location distance data of abnormal time, and obtaining the abnormal indicator analysis results is as follows;

[0053] By utilizing the results of abnormal time stamps, the distance between the abnormal lottery prize redemption location and the known prize redemption location can be calculated. This calculation involves Geographic Information System (GIS) technology or a simple geographic distance formula, such as using the Haversian formula to calculate the distance between two points on the Earth's surface. By summarizing these distance data, spatial analysis can be performed on the locations where abnormal times occur, thereby identifying abnormal patterns or trends. This data will be further used to monitor and prevent potential abnormal behaviors, resulting in abnormal indicator analysis results.

[0054] Please seeFigure 4 According to the abnormal index analysis result, the deviation of the individual prize redemption frequency from the overall average prize redemption frequency is calculated, the data points exceeding two standard deviations are identified as risk transactions, and the frequency deviation statistical result is obtained. The specific steps are as follows:

[0055] S301: Based on the abnormal index analysis result, the lottery redemption frequency of each individual is extracted, the average redemption frequency and the standard deviation of the entire data set are calculated, and the execution process of generating the baseline redemption frequency statistical result is as follows:

[0056] The average redemption frequency and the standard deviation of the entire data set are calculated according to the formula and , the average redemption frequency and the standard deviation are calculated. In the formula, represents the average redemption frequency, represents the standard deviation, represents the redemption frequency of the th individual, is the total number of individuals. The formula details and formula calculation derivation process: set the redemption frequency in the data set as: [3, 5, 2, 8, 3].

[0057] The average redemption frequency is calculated as: .

[0058] The standard deviation is calculated as:

[0059]

[0060]

[0061]

[0062] The results show that the average redemption frequency is 4.2 times, and the standard deviation is 2.1, which is used for subsequent risk assessment.

[0063] S302: Using the baseline redemption frequency statistical result, compare the redemption frequency of each individual with the average redemption frequency, calculate the deviation value, identify the data points whose deviation exceeds two standard deviations, and generate the execution process of the initialization risk identification result as follows:

[0064] Using the baseline redemption frequency statistical result, compare the redemption frequency of each individual with the average redemption frequency, calculate the deviation value of each data point, and set the deviation exceeding two standard deviations as the standard for risk identification. This approach is based on the assumption of normal distribution, in which most data should be within two standard deviations of the mean, and data outside this range is considered potentially abnormal or risky. The calculation in this stage not only reflects the difference between individual and group behavior, but also reveals atypical behavior, generating the initialization risk identification result.

[0065] S303: Extract the data points of the initialized risk identification, mark the point data as risk transactions, analyze the risk level, and analyze the frequency deviation. The execution process of obtaining the frequency deviation statistical result is as follows:

[0066] Extract the data points of the initialized risk identification, mark the point data as risk transactions, analyze the risk level, and analyze the frequency deviation. This includes calculating the deviation of the prize redemption frequency of each risk data point from the average prize redemption frequency, and the distribution characteristics of these deviations, for example, by calculating the mean and variance of the deviation to evaluate the severity of the risk, which helps to identify those who frequently redeem prizes involving abnormal individuals, which can be used to optimize risk control strategies and anti-abnormal measures, and provide practical reference data for decision makers, and obtain the frequency deviation statistical result.

[0067] Please refer to Figure 5 , adopt the frequency deviation statistical result, combine the abnormal index analysis result, create decision tree input data through data integration, execute the decision tree training process, determine the classification threshold through cross-validation, distinguish potential abnormal and normal redemption behaviors, and the steps of building the classification model are as follows:

[0068] S401: Based on the frequency deviation statistical result, combine the abnormal index analysis result, sort and merge the associated data fields, including the redemption time deviation, location distance and individual redemption frequency deviation, and the execution process of generating the decision tree input data is as follows:

[0069] Based on the frequency deviation statistical result, combine the abnormal index analysis result, sort and merge the associated data fields, the process involves integrating the redemption time deviation, location distance and individual redemption frequency deviation as feature variables together to create the input data set of the decision tree model. When merging data, each feature needs to be appropriately converted and standardized to ensure the consistency and comparability of the data, for example, all time deviations are converted to hours, location distances are unified to kilometers, and redemption frequency deviations need to be normalized to facilitate the model to effectively learn the relationship between features, and generate decision tree input data.

[0070] S402: Use the decision tree input data to train the model, set different split parameters and conditions, perform multiple training iterations, optimize the model generalization ability, and the execution process of generating the decision tree training result is as follows:

[0071] By setting different split parameters and conditions, performing multiple training iterations, and according to the formula , the model generalization ability is optimized. In the formula, represents the information gain of the model, is the actual class label, is the class label predicted by the model, is the number of samples. Formula details and formula calculation derivation process: set the model training data as follows: class label , predicted label , calculate the information gain of the model:

[0072]

[0073]

[0074]

[0075] This calculation reflects the error level of the current decision tree model and is used to guide model optimization by adjusting the split parameters to reduce error and improve the prediction accuracy of the model.

[0076] S403: According to the decision tree training result, cross-validation is applied to test and determine the classification threshold of the decision tree model, by comparing the classification effects under different threshold values, the optimal classification threshold is selected to distinguish potential abnormal and normal redemption behaviors, and the execution process of the classification model is as follows:

[0077] According to the decision tree training result, cross-validation is applied to test and determine the classification threshold of the decision tree model, this process selects the optimal classification threshold by comparing the classification effects under different threshold values, using cross-validation technology can avoid overfitting of the model, and ensure the universal applicability of the selected threshold, such as, through ROC curve analysis to evaluate the true positive rate and false positive rate under different threshold values, so as to find out the point that balances the two best, the selected classification threshold is crucial for distinguishing potential abnormal and normal redemption behaviors, and the classification model is obtained.

[0078] Please refer to Figure 6 , using the classification model, analyzing the data of the current redemption request, calculating the abnormal probability of each transaction, and comparing the probability with the preset risk threshold, identifying and marking the transactions exceeding the threshold, and generating the abnormal probability score table, the steps are as follows:

[0079] S501: Through the classification model, load the current redemption request data, and use the constructed classification model to analyze the data, calculate the abnormal time and place of each transaction, and generate the execution process of the transaction abnormal data as follows:

[0080] Using the constructed classification model to analyze the data, according to the formula , calculate the abnormal probability of each transaction. In the formula, represents the calculated abnormal probability, is the input feature of each transaction, is the weight of the corresponding feature in the model, is the number of features. Formula details and formula calculation derivation process:

[0081] Set the input features of a transaction as: [time deviation = 3 hours, location distance = 150 kilometers, redemption frequency deviation = 2 times], and the weight distribution as: [time deviation weight = 0.3, location distance weight = 0.5, redemption frequency deviation weight = 0.2], calculate the abnormal probability: . This calculation shows that the abnormal probability score of this transaction is 76.3, indicating that the transaction has a high risk of abnormality.

[0082] S502: Use transaction abnormal data, use random forest algorithm, evaluate abnormal probability by building multiple decision trees and integrating prediction results, and the execution process of generating abnormal transaction identification results is as follows;

[0083] The formula of random forest algorithm is as follows:

[0084]

[0085] Among them, represents the calculated abnormal probability, represents the number of decision trees, represents the prediction result of the th decision tree, is a preset risk threshold, is an adjustment coefficient, is a natural constant.

[0086] Formula parameter definition and data acquisition

[0087] : The number of decision trees, which can be obtained through model training logs. In this model, 100 decision trees are used.

[0088] : The prediction result of the th decision tree, the specific value is the abnormal probability obtained by applying the decision tree to the validation data set in the model test stage, for example, the prediction probability of a transaction is 0.65.

[0089] : The preset risk threshold, which is determined according to historical data analysis and business needs. This value represents transactions above the probability threshold as high risk. Here, 0.5 is taken as the threshold value.

[0090] : Adjustment coefficient, controls the sensitivity of the model to the prediction results near the threshold, determined by data sensitivity test, set to 1.5 here, to ensure that the model is sensitive to prediction results close to the threshold.

[0091] A specific calculation example

[0092] Setting , , , and in a specific case, 50 decision tree prediction results are 0.7, and the remaining 50 are 0.4.

[0093] Single Calculation steps:

[0094] For :

[0095] For :

[0096] Overall Calculation:

[0097]

[0098] Result interpretation:

[0099] This result shows that the average abnormal probability is 0.1265, which is lower than the preset risk threshold of 0.5, indicating that according to the current model and parameter settings, this transaction is a low-risk transaction. By adjusting values and refining calculations, the model's response to different risk levels can be more accurately controlled, thereby optimizing risk management strategies.

[0100] S503: Review and mark the non-normal transaction identification results, mark the potential non-normal transactions by analyzing the lottery transaction information and the non-normal probability, and obtain the non-normal probability score table. The execution process is as follows:

[0101] Review and mark the initialization of non-normal transaction identification results. This process is carried out by analyzing lottery transaction information and calculating non-normal probability. The review process includes checking each parameter of each transaction, such as transaction time, location, and frequency data, as well as related information such as transaction method and transaction amount. By comparing these data with the non-normal probability of the classification model, it is determined whether the transaction is marked as non-normal, providing specific quantitative data for risk management. These data are the key to detecting and preventing non-normal activities, and a non-normal probability score table is obtained.

[0102] Referring to Figure 7 , according to the non-normal probability score table, filter the transactions marked as risky, send an alert to the relevant regulatory departments and sales points through the automatic notification function, take iterative review and transaction blocking measures, and obtain the steps of the prize redemption monitoring result as follows:

[0103] S601: Extract risk level data of lottery redemption records from the abnormal probability score table, select transaction records marked as risky using the conditional filtering function, and generate a risk transaction list. The execution process is as follows:

[0104] Select transaction records marked as risky using the conditional filtering function, and calculate the number of risky transactions according to the formula . In the formula, represents the number of risky transactions, is the abnormal probability of the th transaction, is an indicator function that takes the value 1 when , and 0 otherwise, is the threshold value of the abnormal probability, is the total number of transactions. Detailed explanation of the formula and formula calculation derivation process: Set the abnormal probabilities of 10 transactions as follows: [0.15, 0.25, 0.55, 0.45, 0.05, 0.75, 0.65, 0.85, 0.35, 0.10], set the abnormal probability threshold value . Calculate the number of transactions that exceed the threshold value: . This indicates that there are a total of 5 transactions marked as risky, and the data is helpful for further monitoring and management of these high-risk transactions.

[0105] S602: Extract data from the risk transaction list, including transaction number, redemption amount, redemption location, and abnormal probability, send to associated regulatory departments and sales points, issue an alert for risky transactions, and generate an alert sending record. The execution process is as follows:

[0106] Extract data from the risk transaction list, including transaction number, redemption amount, redemption location, and abnormal probability, and send these information to associated regulatory departments and sales points. This process includes developing specific information transmission formats and protocols to ensure accurate transmission and reception of data, sending through email or real-time data interface to trigger the alert system. The alert includes detailed transaction information such as number and potential risk level, allowing recipients to quickly identify and respond to abnormal activities, and generating an alert sending record.

[0107] S603: Based on the alert sending record, the regulatory departments and sales points that receive the alert review the risky transactions based on the received data, take corresponding blocking measures, handle the risky transactions by suspending transactions and requesting iterative verification, and obtain the redemption monitoring result. The execution process is as follows:

[0108] Based on the alarm sending record, the regulatory department and the sales point receiving the alarm conduct a review of the risky transaction according to the received data, the process includes verifying and analyzing each piece of received alarm information to determine authenticity and urgency, the measures taken include suspending the transaction and requesting iterative verification, such as requiring the transaction party to provide more evidence or conducting on-site verification, these steps help prevent irregular behavior and ensure the fairness and safety of the transaction, and the prize redemption monitoring result is obtained.

[0109] Please refer to Figure 8 A lottery redemption system based on time series analysis, the lottery redemption system is used to execute the above lottery redemption method based on time series analysis, the system comprises:

[0110] A data extraction module is configured to extract the timestamp, redemption location record and redemption frequency of the lottery redemption request, convert the extracted data into a data format and standardize it, and generate a standardized data set;

[0111] A time series analysis module is configured to perform time series analysis on the timestamps of the standardized data set, identify abnormal deviations from the typical redemption time, calculate the distance between the redemption location and the known redemption location according to the abnormal deviation of the redemption time, and obtain an abnormal index analysis result; the typical redemption time is a preset value of the redemption time determined according to normal redemption behavior;

[0112] A frequency analysis module is configured to calculate the deviation of individual redemption frequency from the overall average redemption frequency according to the abnormal index analysis result, identify data points that exceed twice the standard deviation as risky transactions, and obtain a frequency deviation statistical result;

[0113] A decision tree modeling module is configured to use the frequency deviation statistical result in combination with the abnormal index analysis result to create decision tree input data through data integration, perform a decision tree training process, determine a classification threshold through cross-validation, distinguish between potential irregular and normal redemption behavior, and build a classification model;

[0114] An irregular monitoring module is configured to use the classification model to analyze the data of the current redemption request, calculate the irregular probability of each transaction, and compare the irregular probability with a preset risk threshold, identify and mark transactions that exceed the threshold, and generate an irregular probability score table;

[0115] A risk notification module is configured to filter transactions marked as risky according to the irregular probability score table, send an alarm to the associated regulatory department and sales point through an automatic notification function, take measures such as iterative review and blocking transactions, and obtain a prize redemption monitoring result.

[0116] The above merely describes the preferred embodiments of the present application, and is not intended to limit the present application in other forms. Any skilled person in the art can modify or change the disclosed technical content into equivalent embodiments with equivalent changes, and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, without departing from the technical solution content of the present application, still falls within the protection scope of the present application.

Claims

1. A lottery ticket redemption method based on time series analysis, characterized by, The method comprises the following steps: extracting the timestamp, redemption location record and redemption frequency of the lottery ticket redemption request, converting the extracted data into a data format and standardizing it to generate a standardized data set; performing time series analysis on the timestamp of the standardized data set, identifying abnormal deviations from the typical redemption time, calculating the distance between the redemption location and the known redemption location according to the abnormal deviations of the redemption time, and obtaining an abnormal index analysis result; the typical redemption time is a redemption time preset value determined according to normal redemption behavior; According to the abnormal index analysis result, calculate the deviation of individual redemption frequency from the overall average redemption frequency, identify data points exceeding twice the standard deviation as risky transactions, and obtain a frequency deviation statistical result; Using the frequency deviation statistical result, combined with the abnormal index analysis result, create decision tree input data through data integration, execute decision tree training process, determine classification threshold through cross-validation, distinguish potential abnormal and normal redemption behavior, and build classification model; Using the classification model, analyze the data of the current redemption request, calculate the abnormal probability of each transaction, and compare the abnormal probability with the preset risk threshold, identify and mark the transactions exceeding the threshold, and generate an abnormal probability score table; According to the abnormal probability score table, filter the transactions marked as risky, send an alarm to the associated regulatory department and sales point through the automatic notification function, take measures such as iterative review and blocking transactions, and obtain a redemption monitoring result; Based on the abnormal index analysis result, extract the lottery ticket redemption frequency of each individual, calculate the average redemption frequency and standard deviation of the entire data set, and generate a baseline redemption frequency statistical result; Using the baseline redemption frequency statistical result, compare each individual redemption frequency with the average redemption frequency, calculate the deviation value, identify data points with a deviation exceeding twice the standard deviation, and generate an initialization risk identification result; Extract the data points of the initialization risk identification result and mark the data points as risky transactions, analyze the risk level, and perform frequency deviation analysis to obtain a frequency deviation statistical result; Based on the frequency deviation statistical result, combined with the abnormal index analysis result, sort and merge the associated data fields, including redemption time deviation, location distance and individual redemption frequency deviation, to generate decision tree input data; Using the decision tree input data for model training, setting different split parameters and conditions, performing multiple training iterations, optimizing model generalization ability, and generating decision tree training result; According to the decision tree training result, apply cross-validation to test and determine the classification threshold of the decision tree model, compare the classification effect under different threshold values, select the optimal classification threshold, distinguish potential abnormal and normal redemption behavior, and obtain a classification model.

2. The time series analysis based lottery ticket redemption method of claim 1, wherein, The standardized data set includes a timestamp of a lottery ticket redemption, a longitude and latitude identification of a redemption location, and a redemption frequency value of each redemptions, the abnormal index analysis result includes a deviation value of each redemption timestamp and an average redemption time, a spatial distance of the redemption location and a common location, and an abnormal score, the frequency deviation statistical result includes a comparison value of the individual redemption frequency and the average redemption frequency, a standard deviation calculation result, and a recognized risk transaction marker, the classification model includes a decision tree model structure trained based on input data, a classification threshold, and a consistency index of model verification, the non-normal probability score table includes a transaction number, a calculated non-normal probability, and a marker of whether each transaction exceeds a risk threshold, and the redemption monitoring result includes a list of marked risk transactions, automatic alarm content issued, and review and prevention measures taken.

3. The time series analysis based lottery ticket redemption method of claim 1, wherein, The steps of extracting the timestamp, redemption location record and redemption frequency of the lottery ticket redemption request, converting the extracted data into a data format and standardizing to generate a standardized data set are as follows: Load and read the lottery ticket redemption picture file, apply image recognition technology to identify and record the timestamp and location information in the picture one by one to obtain initialization data records; Based on the initialization data records, each record is traversed and checked to eliminate records with incorrect format and incomplete data, and for the remaining data, the number of redemptions at each location is counted to generate a location redemption frequency statistical table; The data in the location redemption frequency statistical table is uniformly formatted and scaled, and each data is stored according to the uniform standard format, and a standardized data set is generated.

4. The time series analysis based lottery ticket redemption method of claim 1, wherein, The steps of performing time series analysis on the timestamp of the standardized data set, identifying the abnormal deviation from the typical redemption time, and calculating the distance between the redemption location and the known redemption location according to the abnormal deviation of the redemption time to obtain the abnormal index analysis result are as follows: Extract the timestamp from the standardized data set, serialize each timestamp, convert it to continuous time series data, identify the deviation from the typical redemption time through statistical analysis of the time series data, and generate a time deviation analysis result; Based on the time deviation analysis result, the deviation degree of each data point in the time series data from the typical redemption time is calculated, and the time points with deviation exceeding the set threshold are marked according to the deviation size to obtain an abnormal time marker result; Determine the redemption location corresponding to the abnormal time marker result as an abnormal lottery ticket redemption location, calculate the distance between the abnormal lottery ticket redemption location and the known redemption location, and aggregate the location distance data of the abnormal time to obtain the abnormal index analysis result.

5. The time series analysis based lottery ticket redemption method of claim 1, wherein, The steps of using the classification model to analyze the data of the current redemption request, calculating the non-normal probability of each transaction, and comparing the non-normal probability with the preset risk threshold to identify and mark the transactions exceeding the threshold to generate a non-normal probability score table are as follows: Load the current redemption request data through the classification model, and analyze the data using the constructed classification model to calculate the non-normal time and location of each transaction, and generate transaction non-normal data; The transaction abnormal data is used, a random forest algorithm is used, a plurality of decision trees are constructed, and an abnormal probability is evaluated by synthesizing a prediction result to generate an abnormal transaction identification result; The abnormal transaction identification result is reviewed and marked, potential abnormal transactions are marked by analyzing lottery transaction information and the abnormal probability, and an abnormal probability score table is obtained.

6. The time series analysis based lottery ticket redemption method of claim 5, wherein, The formula of the random forest algorithm is as follows: ； wherein, represents the calculated abnormal probability, represents the number of decision trees, represents the prediction result of the th decision tree, is a preset risk threshold, is a regulation coefficient, is a natural constant.

7. The time series analysis based lottery ticket redemption method of claim 1, wherein, According to the abnormal probability score table, transactions marked as risks are screened, an alarm is sent to the associated regulatory department and sales point through an automatic notification function, iterative review and transaction blocking measures are taken, and a prize redemption monitoring result is obtained. Risk level data of a lottery redemption record is extracted from the abnormal probability score table, a transaction record marked as a risk is selected using a conditional filtering function, and a risk transaction list is generated. Data in the risk transaction list, including a transaction number, a redemption amount, a redemption location, and an abnormal probability, is extracted and sent to the associated regulatory department and sales point, an alarm is sent for a risk transaction, and an alarm sending record is generated. Based on the alarm sending record, the regulatory department and sales point receiving the alarm review the risk transaction according to the received data, take corresponding blocking measures, process the risk transaction by suspending the transaction and requesting iterative verification, and obtain a prize redemption monitoring result.

8. A time series analysis based lottery ticket redemption system, characterized by, The lottery redemption method based on time series analysis according to any one of claims 1-7, the system comprises: A data extraction module for extracting the timestamp of the lottery redemption request, the redemption location record, and the redemption frequency, converting the extracted data into a data format and standardizing it to generate a standardized data set; A time series analysis module for performing time series analysis on the timestamp of the standardized data set, identifying abnormal deviations from the typical redemption time, calculating the distance between the redemption location and the known redemption location according to the abnormal deviation of the redemption time, and obtaining an abnormal index analysis result; the typical redemption time is a preset value of the redemption time determined according to normal redemption behavior; A frequency analysis module for calculating the deviation of individual redemption frequency from the overall average redemption frequency according to the abnormal index analysis result, identifying data points exceeding twice the standard deviation as risk transactions, and obtaining a frequency deviation statistical result; A decision tree modeling module for using the frequency deviation statistical result, combining the abnormal index analysis result, creating decision tree input data through data integration, executing a decision tree training process, determining a classification threshold through cross-validation, distinguishing between potential abnormal and normal redemption behaviors, and constructing a classification model; An abnormal monitoring module for using the classification model to analyze the data of the current redemption request, calculating the abnormal probability of each transaction, and comparing the abnormal probability with a preset risk threshold, identifying and marking transactions exceeding the threshold, and generating an abnormal probability score table; A risk notification module for screening transactions marked as risks according to the abnormal probability score table, sending an alarm to the associated regulatory department and sales point through an automatic notification function, taking iterative review and transaction blocking measures, and obtaining a prize redemption monitoring result.

Images