Collusion behavior identification method and device, computer equipment and storage medium

Abstract

The invention relates to a collusion behavior recognition method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining the data of the electricity buyers in an electricity market, wherein the data of the electricity buyers comprises the number of the electricity buyers, and the declared electricity quantity and declared electricity price of each electricity buyer; based on the e-commerce data, determining a collusion behavior identification index data set corresponding to each group of e-commerce in the e-commerce data, the data set including a declaration electric quantity consistency index, a declaration electricity price consistency index, a quotation relative average value, and an e-commerce collusion label; each group of e-commerce businesses comprises two e-commerce businesses; based on the collusion behavior identification index data set corresponding to each group of e-buyers, combining an AdaBoost-DT algorithm to construct a collusion behavior identification model; and based on the collusion behavior identification model, identifying the to-be-identified e-commerce data to determine a collusion behavior identification result. By adopting the method, the result accuracy of collusion behavior recognition can be improved.

Description

Technical Field

[0001] The present application relates to the field of machine learning technology, and in particular to a method, apparatus, computer device and storage medium for identifying collusion behavior. Background Art

[0002] At present, there are many technical means to identify collusion in the power market. Among them, the cloud model processes data through a specific uncertainty conversion model, the fuzzy Petri net uses the combination of fuzzy logic and Petri net (a mathematical model and graphical tool used to describe and analyze concurrent, asynchronous, distributed and other behaviors in the system) to build an analysis model, and the neural network analysis relies on the connection and learning mechanism between neurons to identify collusion. These technologies have played a certain role in the field of power market supervision, providing a reference for the preliminary screening and judgment of collusion, maintaining the trading order of the power market to a certain extent, and are an important part of the power market supervision technology system.

[0003] However, existing collusion identification technologies still have many defects. Methods such as cloud models, fuzzy Petri nets, and neural network analysis fail to fully explore the close connections within the data when processing data, and it is difficult to capture the key features of collusion, so the results of collusion identification are inaccurate. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment and storage medium for collusion behavior identification that can improve the accuracy of collusion behavior identification results in response to the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a method for identifying collusion behavior, comprising:

[0006] Acquire electricity purchasing data in the electricity market, wherein the electricity purchasing data includes: the number of electricity purchasing companies, the declared electricity quantity and the declared electricity price of each electricity purchasing company;

[0007] Based on the e-commerce purchasing data, determining a collusion behavior identification index data set corresponding to each group of e-commerce purchasing included in the e-commerce purchasing data, wherein the data set includes: a reported electricity quantity consistency index, a reported electricity price consistency index, a bid relative comparison mean, and a e-commerce purchasing collusion label; wherein each group of e-commerce purchasing includes two e-commerce purchasing;

[0008] Based on the collusion behavior identification indicator data set corresponding to each group of e-commerce shopping, combined with the AdaBoost - DT algorithm, a collusion behavior identification model is constructed;

[0009] Based on the collusion behavior identification model, the e-commerce shopping data to be identified is identified to determine the collusion behavior identification result.

[0010] In a second aspect, the present application also provides a device for identifying collusion in the electricity market, comprising:

[0011] An acquisition module is used to acquire the data of electricity buyers in the electricity market, wherein the data of electricity buyers includes: the number of electricity buyers, the declared electricity quantity and the declared electricity price of each electricity buyer;

[0012] A determination module is used to determine, based on the electricity purchaser data, a collusion behavior identification index data set corresponding to each group of electricity purchasers included in the electricity purchaser data, wherein the data set includes: a reported electricity quantity consistency index, a reported electricity price consistency index, a bid relative comparison mean, and an electricity purchaser collusion label; wherein each group of electricity purchasers includes two electricity purchasers;

[0013] A construction module is used to construct a collusion behavior identification model based on the collusion behavior identification indicator data set corresponding to each group of e-commerce shopping platforms and in combination with the AdaBoost - DT algorithm;

[0014] The identification module is used to identify the e-commerce shopping data to be identified based on the collusion behavior identification model to determine the collusion behavior identification result.

[0015] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0016] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in the first aspect.

[0017] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.

[0018] The above-mentioned method for identifying collusion in the power market can obtain the data of power buyers in the power market, and the data of power buyers includes: the number of power buyers, the declared power quantity and the declared power price of each power buyer; based on the data of power buyers, determine the collusion behavior identification index data set corresponding to each group of power buyers included in the power buyer data, and the data set includes: the declared power quantity consistency index, the declared power price consistency index, the relative comparison mean of the quotation, and the collusion label of the power buyers; wherein each group of power buyers includes two power buyers; based on the collusion behavior identification index data set corresponding to each group of power buyers, combined with the AdaBoost - DT algorithm, a collusion behavior identification model is constructed; based on the collusion behavior identification model, the data of power buyers to be identified are identified to determine the collusion behavior identification result. This technical solution constructs a collusion behavior identification index data set by obtaining the data of power buyers in the power market, including the number of power buyers, the declared power quantity and the declared power price, etc. This data set covers key elements such as the declared power quantity consistency index, the declared power price consistency index, the relative comparison mean of the quotation, and the collusion label of the power buyers, and takes each group of two power buyers as the analysis unit, which can accurately focus on the correlation characteristics between the power buyers. By using the AdaBoost - DT algorithm to build a collusion behavior identification model, we can give full play to the advantages of the algorithm, efficiently process and analyze data, effectively integrate multi-dimensional indicator information, and improve the accuracy and generalization ability of the model. Based on this model, the data of the electricity buyers to be identified can be identified quickly and accurately to determine the collusion behavior identification results, which will help the power market regulatory authorities to promptly discover and deal with possible collusion between electricity buyers. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] FIG1 is a schematic flow chart of a method for identifying collusion in a power market in one embodiment;

[0021] Figure 2 is a curve showing the running time versus the number of iterations in an example;

[0022] Figure 3 shows the classification data of the test set in an example;

[0023] FIG4 is a structural block diagram of a device for identifying collusion in the power market in one embodiment;

[0024] FIG. 5 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0026] At present, there are many technical means to identify collusion in the power market. Among them, the cloud model processes data through a specific uncertainty conversion model, the fuzzy Petri net uses the combination of fuzzy logic and Petri net (a mathematical model and graphical tool used to describe and analyze concurrent, asynchronous, distributed and other behaviors in the system) to build an analysis model, and the neural network analysis relies on the connection and learning mechanism between neurons to identify collusion. These technologies have played a certain role in the field of power market supervision, providing a reference for the preliminary screening and judgment of collusion, maintaining the trading order of the power market to a certain extent, and are an important part of the power market supervision technology system.

[0027] However, existing collusion identification technologies still have many defects. Methods such as cloud models, fuzzy Petri nets, and neural network analysis fail to fully explore the close connections within the data when processing data, and it is difficult to capture the key features of collusion, so the results of collusion identification are inaccurate.

[0028] In order to solve the above problems, a method for identifying collusion in the power market is provided. By obtaining the data of power buyers in the power market, including the number of power buyers, declared electricity quantity and declared electricity price, a collusion behavior identification index data set is constructed. This data set covers key elements such as the declared electricity quantity consistency index, the declared electricity price consistency index, the relative mean of the bid price and the collusion label of power buyers. In addition, two power buyers in each group are used as analysis units, which can accurately focus on the correlation characteristics between power buyers. By using the AdaBoost - DT algorithm to build a collusion behavior identification model, the advantages of the algorithm can be fully utilized to efficiently process and analyze the data, effectively integrate multi-dimensional indicator information, and improve the accuracy and generalization ability of the model. Based on this model, the data of power buyers to be identified can be identified, and the collusion behavior identification results can be determined quickly and more accurately, which is helpful for power market regulatory departments to promptly discover and deal with possible collusion between power buyers.

[0029] In an exemplary embodiment, as shown in FIG1 , a flow chart of a method for identifying collusion in the electricity market is provided. The method may include the following steps 101 to 104:

[0030] 101. Obtain data on electricity buyers in the electricity market.

[0031] In the embodiment of the present application, the electricity purchasing data includes: the number of electricity purchasing companies, the declared electricity quantity and the declared electricity price of each electricity purchasing company.

[0032] The electricity purchasing data in the above-mentioned electricity market may be historical data in the electricity market.

[0033] The above step 101 is a data collection process, during which data such as the number of electricity buyers in the electricity market, the amount of electricity applied for by each electricity buyer, and the electricity price declared by each electricity buyer can be collected.

[0034] In some embodiments, after acquiring the electricity purchaser data in the electricity market, the electricity purchaser data may be preprocessed first.

[0035] In some embodiments, after obtaining the electricity purchasing data in the electricity market, data outliers and data missing values ​​can be determined from the electricity purchasing data, and then, the data outliers and data missing values ​​in the electricity purchasing data can be corrected using regression interpolation method to obtain valid electricity purchasing data that can be used for subsequent calculation and analysis.

[0036] Among them, data outliers are measured values ​​in e-commerce data where the deviation from the data mean exceeds two times the standard deviation.

[0037] For example, the regression interpolation method can be combined with the following formula (1) to correct the detected data outliers and missing values ​​by replacing them with the average value of n data before and after the corresponding position.

[0038] (1)

[0039] in, represents the corrected kth missing value / outlier value, It represents the nth observed value before the kth missing value or data outlier in the observed sample. It represents the nth observed value after the kth missing data value or data outlier in the observed sample. n is a constant term, and here n=4 can be taken.

[0040] In the above embodiment, preprocessing the data of the electricity supplier has many significant technical effects. First, by determining the data outliers and missing data values, the bad parts in the data can be accurately identified. Data outliers are often measured values ​​that deviate greatly from the overall data distribution, which may be caused by measurement errors, system failures or special circumstances. If not processed, they will seriously interfere with the accuracy of subsequent analysis. The existence of missing data values ​​will make the data incomplete, affecting the integrity and reliability of model construction and analysis. The use of regression interpolation method to correct these outliers and missing values, based on the regression interpolation formula, combined with a specific number of observations before and after (such as n = 4) to average to replace abnormal or missing data, can effectively restore the integrity and continuity of the data. The effective electricity supplier data obtained in this way can be used for subsequent calculation and analysis, so that the constructed collusion behavior identification model is based on a more accurate and complete data foundation, thereby improving the accuracy of the model in identifying the collusion behavior of electricity suppliers, reducing the misjudgment or omission caused by data problems, and enhancing the reliability and effectiveness of the entire power market collusion behavior identification technology solution, providing more powerful data support and decision-making basis for power market supervision.

[0041] 102. Based on the e-commerce shopping data, determine a collusion behavior identification indicator data set corresponding to each group of e-commerce shopping included in the e-commerce shopping data.

[0042] The data set includes: reported electricity quantity consistency index, reported electricity price consistency index, quotation relative comparison mean, and electricity purchaser collusion label; each group of electricity purchasers includes two electricity purchasers.

[0043] In some embodiments, for each group of electricity buyers, a reported electricity price consistency index may be determined based on a difference between electricity prices reported by two electricity buyers in each time period and an average of electricity prices reported by all electricity buyers in the corresponding time period.

[0044] For example, the consistency index of electricity prices reported by any two power suppliers can be obtained based on the following formula (2).

[0045] (2)

[0046] Among them, in formula (2) Indicates the consistency of the quotations of the two shopping merchants. Indicates Time period, Indicates In this bidding, the individual shopping e-commerce The declared price for the time period, Indicates that the jth shopping e-commerce company is in the first The declared price for the time period. For all the online shopping merchants in this auction The average price of the period. When the value of is close to 0, the buyers are more likely to make consistent bids, which means there may be collusion.

[0047] In the embodiment corresponding to the above formula (2), the reported electricity price consistency index is determined by constructing a difference relationship between the electricity prices reported by two electricity buyers in each period and the average electricity prices reported by all electricity buyers in the corresponding period. The technical effect is that the consistency of the quotations between any two electricity buyers can be quantitatively measured. The closer the value is to 0, the more consistent the electricity prices reported by the two power buyers are. From this, it can be inferred that the power buyers are more likely to make consistent quotations, which helps to determine whether there is collusion. This provides an intuitive and quantifiable analysis basis for accurately identifying collusion among power buyers in the power market, helps to improve the accuracy and scientificity of monitoring and identifying collusion in the power market, and ensure the fairness and justice of the power market trading order.

[0048] In some embodiments, for each group of electricity purchasers, a reported electricity consistency index can be determined based on the difference relationship between the electricity reported by two electricity purchasers in each time period and the average electricity reported by all electricity purchasers in the corresponding time period.

[0049] For example, the reported electricity quantity consistency index of any two power suppliers can be obtained based on the following formula (3).

[0050] (3)

[0051] in, Indicates the consistency of the electricity reported by the two electricity buyers. Indicates Time period, , are the i-th and j-th shopping e-commerce merchants in this bidding. The reported electricity consumption during the period, For all the online shopping merchants in this auction The average value of electricity consumption reported during a period. The closer it is to 0, the more likely it is that electricity buyers will consistently apply for electricity, which means there may be collusion.

[0052] In the embodiment corresponding to the above formula (3), the reported power consistency index is determined by quantifying the difference between the power reported by two electricity buyers in each time period and the average power reported by all electricity buyers in the corresponding time period. This method can accurately measure the consistency of the power reported by any two electricity buyers. This indicator can be used as an effective judgment tool. The closer it is to 0, the more consistent the two power buyers are in declaring their electricity. This provides an intuitive and quantifiable basis for identifying whether there is collusion between power buyers. In the context of power market supervision, this indicator can be used to monitor the behavior of power buyers more scientifically and accurately, and help to promptly discover collusion that may manipulate the market through consistent declaration of electricity, thereby maintaining the fairness and impartiality of the power market declaration process and ensuring the healthy and orderly operation of the power market.

[0053] In some embodiments, for each group of online shopping merchants, a relative comparison mean of quotations may be determined based on the respective average quotations of two online shopping merchants and the total average quotations of all online shopping merchants.

[0054] For example, the relative comparison mean of the quotations of any two power suppliers can be obtained based on the following formula (4).

[0055] (4)

[0056] In formula (4), It represents the relative mean of the quotations of two online shopping platforms. Indicates The average electricity price reported by individual electricity buyers, represents the average electricity price reported by the jth electricity buyer. The further away from 100%, the greater the possibility of collusion.

[0057] In the embodiment related to the above formula (4), the relative comparison mean of the quotations is determined based on the average quotations of the two electricity buyers and the total average quotations of all electricity buyers, thus constructing a unique evaluation dimension. Its technical effect is remarkable, and the indicator can measure the relationship between electricity buyers from the perspective of relative quotation ratio. When X_3ij is further away from 100%, it can effectively indicate that the possibility of collusion among electricity buyers is greater, which provides another powerful basis for judgment for power market supervision. In synergy with the reported power consistency index, the behavior pattern of electricity buyers is comprehensively analyzed from different aspects, further improving the accuracy and reliability of collusion behavior identification. It can more keenly capture the abnormal coordination situation that may exist in the quotation link of electricity buyers, effectively assist the regulatory authorities to accurately locate potential collusion, ensure the standardization and transparency of the quotation link of the power market, promote the transaction of power resources in a fair and just environment, promote the stable and efficient operation of the power market mechanism, and reduce the market risks and resource mismatch caused by collusion.

[0058] In some embodiments, after determining the above-mentioned reported electricity quantity consistency index, reported electricity price consistency index, and quotation relative mean, a buyer-seller collusion label can also be set, and the reported electricity quantity consistency index, reported electricity price consistency index, quotation relative mean, and buyer-seller collusion label can be integrated into a collusion behavior identification indicator data set.

[0059] Among them, the e-commerce collusion label indicates no collusion or collusion.

[0060] For example, the collusion behavior identification indicator dataset can be expressed as shown in the following formula (5):

[0061] = (5)

[0062] in, represents the sequence of indicators of the collusion behavior identification of two online shopping merchants. That is the The e-commerce collusion label between the first e-commerce buyer and the jth e-commerce buyer, ∈{-1,1},-1 means no collusion, 1 means collusion.

[0063] In some embodiments, the e-commerce collusion tag can be set to collusion in target situations, including but not limited to:

[0064] (1) It means that the quotations of the two e-sellers are the same, that is, The difference between the value and 0 is less than a first preset value;

[0065] (2) It means that the electricity reported by the two electricity buyers is the same, that is, The difference between the value and 0 is less than a second preset value;

[0066] (3) It means that the quotations of the two online shopping platforms are far from 100% of the mean, that is, The difference between the percentage and 100% is greater than a third preset value.

[0067] Among them, the above-mentioned first preset value, second preset value and third preset value can be set according to actual needs, and the embodiment of the present application is not limited thereto.

[0068] The above formula (5) integrates the reported electricity quantity consistency index, reported electricity price consistency index, bid relative comparison mean and electricity purchaser collusion label into a data set, constructing a comprehensive and intuitive judgment basis. By combining various indicators with collusion labels, it can comprehensively present the relationship between electricity purchasers, facilitate more accurate identification and judgment of whether there is collusion among electricity purchasers, and provide strong support for electricity market supervision.

[0069] In some embodiments, after obtaining the above-mentioned collusion behavior identification indicator data set, the data set may also be subjected to data normalization processing.

[0070] For example, it can be shown as the following formula (6):

[0071] (6)

[0072] 103. Based on the collusion behavior identification indicator data set corresponding to each group of e-commerce shopping, combined with the AdaBoost-DT algorithm, a collusion behavior identification model is constructed.

[0073] Among them, the AdaBoost - DT algorithm is an algorithm used to build a collusion behavior identification model. It is a method of combining multiple weak classifiers into a strong classifier. When processing data, this algorithm will first build many simple classifiers based on existing data features (such as declared electricity consistency indicators, declared electricity price consistency indicators, quotation relative comparison mean, etc.). These simple classifiers may not work very well when used alone, just like a small judgment tool that is not very accurate. However, through this algorithm, different weights will be assigned to these classifiers. In the process of continuous iteration, the classifiers corresponding to the data features that can be more accurately classified will play a greater role, that is, the weight is higher. Finally, these weighted classifiers are combined to form a comprehensive and more powerful classifier, so as to make a more accurate judgment on whether there is collusion among e-commerce shopping companies. This algorithm can make full use of various feature information in the data and effectively improve the accuracy of the model in identifying collusion.

[0074] In some embodiments, based on the collusion behavior identification indicator dataset corresponding to each group of e-commerce shopping, combined with the AdaBoost - DT algorithm, a collusion behavior identification model is constructed, including but not limited to the following steps:

[0075] Step 1: Based on the collusion behavior identification indicator dataset corresponding to each group of e-shopping merchants, determine the training set and the dataset; the sample weight of each sample initially set in the training set is the same.

[0076] Among them, this step is mainly the preparation work of data division. Based on the collusion behavior identification indicator data set corresponding to the e-commerce shopping, it is divided into two parts: the training set and the data set. In the initial stage of the training set, each sample is given the same weight, which can ensure that each sample has an equal opportunity to be learned at the beginning of subsequent training, laying the foundation for the basic data distribution for building an accurate model.

[0077] Step 2: Train a basic decision tree classifier based on the training set, and determine the training error rate of the basic decision tree classifier.

[0078] Among them, the divided training set is used to train the basic decision tree classifier. During the training process, the decision tree structure is constructed by learning and analyzing the data features. Then, the training error rate of the basic decision tree classifier is calculated according to the difference between the training results and the actual labels. The error rate can intuitively reflect the accuracy of the current basic decision tree classifier and is a key indicator for the subsequent evaluation and improvement of the classifier.

[0079] For example, the input data set and sample weights can be initialized: 1) Assuming that the number of samples in the data set is N, the initial weight of each sample is set to 1 / N; 2) The base classifier is trained using the training sample set , calculated The training error rate is then The weight of The training error rate can be expressed as the following formula (7):

[0080] (7)

[0081] in, express The training error rate is Respectively represent the data set sequence of e-commerce shopping, N represents the total number of e-commerce shopping, if If it is greater than 50%, the base classifier is discarded.

[0082] Step 3: Based on the training error rate, determine the importance score of a basic decision tree classifier.

[0083] Among them, according to the training error rate obtained in the previous step, the importance score of the basic decision tree classifier is further determined. A classifier with a low error rate indicates that it has a high accuracy in data classification, so it is highly important in the entire model construction process, and its importance score is also correspondingly high; conversely, a classifier with a high error rate has a low importance score, which helps to screen out more valuable classifiers for combination in the future.

[0084] The importance score of the above-mentioned basic decision tree classifier may be the weight of the basic decision tree classifier. Exemplarily, the weight of the basic decision tree classifier may be shown in the following formula (8):

[0085] = (8)

[0086] in, express The weight of express The training error rate.

[0087] Step 4: Based on the training error rate, update the sample weights of the samples in the training set to determine the updated training set.

[0088] Among them, the weights of the samples in the training set are updated again based on the training error rate, so that the weights of the samples that are misclassified in this classification are increased, so that in subsequent training cycles, the classifier can pay more attention to these error-prone samples, and then obtain an updated training set, making the training process more targeted and effective, and continuously optimizing the classifier performance.

[0089] After looping through steps 2 to 4, the multiple basic decision tree classifiers are linearly combined based on the importance score of each basic decision tree classifier to obtain a strong classifier, which is used as a collusion behavior recognition model.

[0090] In the above embodiment, the collusion behavior identification model is constructed by the AdaBoost-DT algorithm, which can fully mine the information in the collusion behavior identification indicator data set. The training set and the data set are divided and the initial weight is assigned to ensure equal learning opportunities for the data. The basic decision tree classifier is trained and the error rate is calculated to accurately evaluate its performance. The importance score is determined according to the error rate to facilitate the screening of high-quality classifiers. The sample weight is updated to make the training more targeted. After cyclic iteration, the linear combination of multiple classifiers forms a strong classifier, which effectively improves the accuracy and reliability of the model in identifying the collusion behavior of electricity buyers and effectively maintains the order of the electricity market.

[0091] In some embodiments, steps 2 to 4 are executed in a loop, and multiple basic decision tree classifiers are linearly combined based on the importance score of each basic decision tree classifier to obtain a strong classifier, including: executing steps 2 to 4 in a loop, calculating the overall error rate of the current training, where the overall error rate is determined based on the training error rate of all current basic decision tree classifiers; when the overall error rate is less than or equal to a preset error rate and the number of iterations is greater than or equal to a preset number, the multiple basic decision tree classifiers are linearly combined based on the importance score of each basic decision tree classifier to obtain a strong classifier.

[0092] In the above embodiment, a strong classifier is obtained by cyclically training the basic decision tree classifier and performing linear combination according to specific conditions. The overall error rate is calculated during the cycle, and the comprehensive performance of all current basic decision tree classifiers can be comprehensively measured. It is determined based on the training error rate of each classifier and can accurately reflect the overall deviation degree of the model. The preset error rate and the preset number of times are used as termination conditions to ensure that the model meets certain standards in accuracy and stability. When the overall error rate is less than or equal to the preset error rate and the number of iterations meets the standard, a linear combination is performed, which not only avoids the risk of overfitting, but also ensures sufficient training and optimization of the model. The strong classifier obtained in this way can more accurately identify the collusion behavior of electricity buyers, effectively improve the reliability and adaptability of the collusion behavior identification model in a complex power market environment, provide strong technical guarantees for power market supervision, and maintain the fairness, justice and orderly operation of market transactions.

[0093] In some embodiments, when executing steps 2 to 4 for the first time, if it is determined that the training error rate determined in step 2 is less than 50%, continue to execute steps 3 and 4; if it is determined that the training error rate determined in step 2 is greater than or equal to 50%, delete a basic decision tree classifier, and re-execute step 2 to train a new basic decision tree classifier until the training error rate is less than 50%, and then continue to execute steps 3 and 4.

[0094] 104. Based on the collusion behavior identification model, the e-commerce shopping data to be identified is identified to determine the collusion behavior identification results.

[0095] In the above embodiment, when steps 2 to 4 are executed for the first time, the training error rate is compared with 50%. If it is greater than or equal to 50%, the current basic decision tree classifier is deleted and retrained. This measure can effectively exclude those basic classifiers that perform extremely poorly in the initial stage and are almost impossible to accurately classify, avoid their negative impact on subsequent model construction, and ensure that the basic decision tree classifiers involved in model construction have the most basic accuracy and effectiveness. When the training error rate is less than 50%, the subsequent steps will be continued, which will help to screen out relatively more reliable and more potential basic classifiers for further optimization and combination, so that the collusion behavior identification model constructed by the entire AdaBoost-DT algorithm has a higher quality in the basic components, thereby laying a solid foundation for finally obtaining a high-precision, high-reliability strong classifier, improving the accuracy and stability of the identification of collusion behavior of electricity buyers, and ensuring the effectiveness and fairness of electricity market supervision.

[0096] For example, an example analysis is conducted in combination with the specific transaction data of 25 electricity buyers in a certain province's electricity market. As shown in FIG2 , the curve of the running time changing with the number of iterations in the example is shown. FIG3 shows the classification data of the test set in the example. The steps of applying the method provided in the embodiment of the present application in this example are as follows:

[0097] a. Collect data. According to the model requirements, collect data such as the number of electricity buyers in the market, the electricity volume reported by electricity buyers, and the electricity prices reported by electricity buyers. The collection period is one month, the data collection frequency is 1 hour, and there are 24 time nodes a day. According to the model prediction requirements, the processed data is divided into training set and test set in a ratio of 7:3.

[0098] b. Data processing: take n=4 and correct the data outliers and missing values ​​in the data according to the above formula (1).

[0099] c. Based on the collected historical data, a data set of collusion behavior identification indicators corresponding to each group of e-commerce shopping platforms is established. The above formula is used to , formula (3) and formula (4) are used to obtain the three monitoring index values ​​of quotation consistency, declared quantity consistency and quotation relative comparison mean of any two suppliers, forming an identification index data set, and the data is normalized according to the above formula (6). These three indicators are used to comprehensively identify whether any two online shopping merchants in the market have engaged in collusion.

[0100] d. Establish a collusion behavior identification model, iterate repeatedly according to formula (7) and formula (8) to obtain the trained collusion behavior identification model, and record the parameter values. Finally, bring the power generator data to be identified into the collusion behavior identification model to obtain the result value. The above parameter values ​​can be: L=60, ε=0.011%, that is, the entire model has undergone 60 iterations in the training process, and the error rate is 0.011%.

[0101] The method for identifying collusion in the electricity market shown in the embodiment of the present application obtains the data of electricity buyers in the electricity market, including the number of electricity buyers, the declared electricity quantity and the declared electricity price, and constructs a collusion identification index data set. This data set covers key elements such as the declared electricity quantity consistency index, the declared electricity price consistency index, the relative mean of the quotations and the collusion label of electricity buyers, and takes each group of two electricity buyers as the analysis unit, which can accurately focus on the correlation characteristics between the two electricity buyers. By constructing a collusion identification model with the help of the AdaBoost-DT algorithm, the advantages of the algorithm can be fully utilized to efficiently process and analyze the data, effectively integrate multi-dimensional indicator information, and improve the accuracy and generalization ability of the model. Based on this model, the data of the electricity buyers to be identified can be identified, and the collusion identification results can be determined quickly and more accurately, which is helpful for the electricity market regulatory authorities to promptly discover and deal with possible collusion between electricity buyers.

[0102] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0103] Based on the same inventive concept, the embodiment of the present application also provides a collusion behavior identification device for implementing the collusion behavior identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more collusion behavior identification device embodiments provided below can refer to the limitations of the collusion behavior identification method above, and will not be repeated here.

[0104] In an exemplary embodiment, as shown in FIG4 , a collusion behavior identification device is provided, including:

[0105] The acquisition module 401 is used to acquire the data of electricity buyers in the electricity market, wherein the data of electricity buyers includes: the number of electricity buyers, the declared electricity quantity and the declared electricity price of each electricity buyer;

[0106] The determination module 402 is used to determine, based on the e-commerce purchaser data, a collusion behavior identification index data set corresponding to each group of e-commerce purchasers included in the e-commerce purchaser data, wherein the data set includes: a reported power consistency index, a reported electricity price consistency index, a bid relative comparison mean, and a e-commerce purchaser collusion label; wherein each group of e-commerce purchasers includes two e-commerce purchasers;

[0107] A construction module 403 is used to construct a collusion behavior identification model based on the collusion behavior identification indicator data set corresponding to each group of e-shopping merchants in combination with the AdaBoost - DT algorithm;

[0108] The identification module 404 is used to identify the e-shopping data to be identified based on the collusion behavior identification model to determine the collusion behavior identification result.

[0109] In some embodiments, the determining module 402 is specifically configured to:

[0110] For each group of e-commerce merchants:

[0111] Based on the difference relationship between the power reported by two electricity buyers in each time period and the average power reported by all electricity buyers in the corresponding time period, the difference relationship is determined to determine the reported power consistency index;

[0112] Determine the consistency index of the declared electricity price based on the difference between the electricity prices declared by the two electricity buyers in each time period and the average electricity prices declared by all electricity buyers in the corresponding time period;

[0113] Determine the relative comparison mean of the quotations based on the average quotations of the two online buyers and the total average quotations of all online buyers;

[0114] Setting a collusion label for e-commerce purchasers, wherein the collusion label for e-commerce purchasers indicates no collusion or collusion;

[0115] The reported electricity quantity consistency index, the reported electricity price consistency index, the quotation relative comparison mean and the electricity purchasing company collusion label are integrated into the collusion behavior identification index data set.

[0116] In some embodiments, the determining module 402 is further configured to:

[0117] Determine data outliers and data missing values ​​from the e-commerce shopping data, wherein the data outliers are measured values ​​in the e-commerce shopping data whose deviation from the data mean exceeds two times the standard deviation;

[0118] The data outliers and missing values ​​in the e-commerce shopping data are corrected by using the regression interpolation method to obtain valid e-commerce shopping data that can be used for subsequent calculation and analysis.

[0119] In some embodiments, the construction module 403 is further used to:

[0120] Step 1: Based on the collusion behavior identification indicator data sets corresponding to the groups of e-shopping merchants, a training set and a data set are determined; the sample weight of each sample initially set in the training set is the same;

[0121] Step 2: training a basic decision tree classifier based on the training set, and determining the training error rate of the basic decision tree classifier;

[0122] Step 3: Determine the importance score of the basic decision tree classifier based on the training error rate;

[0123] Step 4: Based on the training error rate, update the sample weights of the samples in the training set to determine the updated training set;

[0124] After looping through steps 2 to 4, the multiple basic decision tree classifiers are linearly combined based on the importance score of each basic decision tree classifier to obtain a strong classifier, and the strong classifier is used as the collusion behavior identification model.

[0125] In some embodiments, the construction module 403 is specifically used to:

[0126] Execute steps 2 to 4 in a loop to calculate the overall error rate of the current training, where the overall error rate is determined based on the training error rates of all current basic decision tree classifiers;

[0127] When the overall error rate is less than or equal to the preset error rate and the number of iterations is greater than or equal to the preset number, the multiple basic decision tree classifiers obtained are linearly combined based on the importance score of each basic decision tree classifier to obtain a strong classifier.

[0128] In some embodiments, the construction module 403 is specifically used to:

[0129] When executing steps 2 to 4 for the first time, if it is determined that the training error rate determined in step 2 is less than 50%, continue to execute steps 3 and 4; if it is determined that the training error rate determined in step 2 is greater than or equal to 50%, delete the basic decision tree classifier, and re-execute step 2 to train a new basic decision tree classifier until the training error rate is less than 50%, and then continue to execute steps 3 and 4.

[0130] Each module in the above-mentioned collusion behavior identification device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0131] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be shown in FIG5. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for identifying collusion behavior is implemented.

[0132] Those skilled in the art will understand that the structure shown in FIG. 5 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0133] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the method for identifying collusion shown in the above embodiment is implemented.

[0134] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for identifying collusion behavior shown in the above embodiment is implemented.

[0135] In one embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the method for identifying collusion behavior shown in the above embodiment.

[0136] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0137] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0138] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying collusion, characterized in that: The method comprises: Acquire electricity purchasing data in the electricity market, wherein the electricity purchasing data includes: the number of electricity purchasing companies, the declared electricity quantity and the declared electricity price of each electricity purchasing company; Based on the e-commerce purchasing data, determining a collusion behavior identification index data set corresponding to each group of e-commerce purchasing included in the e-commerce purchasing data, wherein the data set includes: a reported power consistency index, a reported electricity price consistency index, a bid relative comparison mean, and a e-commerce purchasing collusion label; each group of e-commerce purchasing includes two e-commerce purchasing; Based on the collusion behavior identification indicator data set corresponding to each group of e-commerce shopping, combined with the AdaBoost - DT algorithm, a collusion behavior identification model is constructed; Based on the collusion behavior identification model, the e-commerce shopping data to be identified is identified to determine the collusion behavior identification result.

2. The method according to claim 1, characterized in that The step of determining, based on the e-shopping data, a collusion behavior identification indicator data set corresponding to each group of e-shopping data included in the e-shopping data, comprises: For each group of e-commerce merchants: Based on the difference relationship between the power reported by two electricity buyers in each time period and the average power reported by all electricity buyers in the corresponding time period, the difference relationship is determined to determine the reported power consistency index; Determine the consistency index of the declared electricity price based on the difference between the electricity prices declared by the two electricity buyers in each time period and the average electricity prices declared by all electricity buyers in the corresponding time period; Determine the relative comparison mean of the quotations based on the average quotations of the two online buyers and the total average quotations of all online buyers; Setting a collusion label for e-commerce purchasers, wherein the collusion label for e-commerce purchasers indicates no collusion or collusion; The reported electricity quantity consistency index, the reported electricity price consistency index, the quotation relative comparison mean and the electricity purchasing company collusion label are integrated into the collusion behavior identification index data set.

3. The method according to claim 1, characterized in that After acquiring the electricity purchaser data in the electricity market, and before determining, based on the electricity purchaser data, a collusion behavior identification indicator data set corresponding to each group of electricity purchasers included in the electricity purchaser data, the method further includes: Determine data outliers and data missing values ​​from the e-commerce shopping data, wherein the data outliers are measured values ​​in the e-commerce shopping data whose deviation from the data mean exceeds two times the standard deviation; The data outliers and missing values ​​in the e-commerce shopping data are corrected by using the regression interpolation method to obtain valid e-commerce shopping data that can be used for subsequent calculation and analysis.

4. The method according to claim 1, characterized in that: The collusion behavior identification model is constructed based on the collusion behavior identification indicator data set corresponding to each group of e-commerce shopping platforms and combined with the AdaBoost - DT algorithm, including: Step 1: Based on the collusion behavior identification indicator data sets corresponding to the groups of e-shopping merchants, a training set and a data set are determined; the sample weight of each sample initially set in the training set is the same; Step 2: training a basic decision tree classifier based on the training set, and determining a training error rate of the basic decision tree classifier; Step 3: Determine the importance score of the basic decision tree classifier based on the training error rate; Step 4: Based on the training error rate, update the sample weights of the samples in the training set to determine the updated training set; After looping through steps 2 to 4, the multiple basic decision tree classifiers are linearly combined based on the importance score of each basic decision tree classifier to obtain a strong classifier, and the strong classifier is used as the collusion behavior identification model.

5. The method according to claim 4, characterized in that The loop executes step 2 to step 4, linearly combines the obtained multiple basic decision tree classifiers based on the importance score of each basic decision tree classifier to obtain a strong classifier, including: Execute steps 2 to 4 in a loop to calculate the overall error rate of the current training, where the overall error rate is determined based on the training error rates of all current basic decision tree classifiers; When the overall error rate is less than or equal to the preset error rate and the number of iterations is greater than or equal to the preset number, the multiple basic decision tree classifiers obtained are linearly combined based on the importance score of each basic decision tree classifier to obtain a strong classifier.

6. The method according to claim 5, characterized in that When executing steps 2 to 4 for the first time, if it is determined that the training error rate determined in step 2 is less than 50%, continue to execute steps 3 and 4; if it is determined that the training error rate determined in step 2 is greater than or equal to 50%, delete the basic decision tree classifier, and re-execute step 2 to train a new basic decision tree classifier until the training error rate is less than 50%, and then continue to execute steps 3 and 4.

7. A device for identifying collusion behavior, characterized in that: The device comprises: An acquisition module is used to acquire the data of electricity buyers in the electricity market, wherein the data of electricity buyers includes: the number of electricity buyers, the declared electricity quantity and the declared electricity price of each electricity buyer; A determination module is used to determine, based on the electricity purchaser data, a collusion behavior identification index data set corresponding to each group of electricity purchasers included in the electricity purchaser data, wherein the data set includes: a reported electricity quantity consistency index, a reported electricity price consistency index, a bid relative comparison mean, and an electricity purchaser collusion label; wherein each group of electricity purchasers includes two electricity purchasers; A construction module is used to construct a collusion behavior identification model based on the collusion behavior identification indicator data set corresponding to each group of e-commerce shopping platforms and in combination with the AdaBoost-DT algorithm; The identification module is used to identify the e-commerce shopping data to be identified based on the collusion behavior identification model to determine the collusion behavior identification result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6.

Images