Big data analysis model evaluation method for market research and judgment and risk management

Abstract

The invention discloses a big data analysis model evaluation method for market research and judgment and risk management, and relates to the field of data processing, and the method comprises the steps: constructing and forming a historical crisis event database; obtaining an evaluation result; obtaining at least one evaluation defect; forming an evaluation fault point based on the evaluation defect; obtaining the importance coefficient of the evaluation fault point; obtaining a to-be-tested data set; obtaining a to-be-tested data block; carrying out weight endowing on the to-be-tested data block; and calculating a data evaluation perfection coefficient of the to-be-tested data block to obtain a total evaluation perfection coefficient of the big data analysis model. An evaluation fault point is formed based on an evaluation defect, an important coefficient of the evaluation fault point is obtained, a to-be-tested data set is obtained, a data evaluation perfection coefficient of a to-be-tested data block is calculated, and a big data analysis model is evaluated from two aspects of data identification blind areas and data processing path missing. And the evaluation condition of the big data analysis model can be truly reflected.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a big data analysis model evaluation method for market research and risk management. Background Art

[0002] With the increasing development of digital governance and intelligent applications, the challenges associated with the analysis and application of large-scale data models are becoming increasingly prominent. For large-scale optimization problems in reality, the performance of large-scale data models decreases dramatically as the number of dimensions of the optimized target increases. These application issues pose a significant challenge to the development of modern industries.

[0003] Since existing big data analysis models need to deal with the analysis of a large amount of uncertain data, their analysis results are of great significance to the validity of the data. In the existing analysis process, due to the limitations of big data analysis models, data analysis is easily missing, but the existing big data analysis model evaluation does not adequately assess the situation of missing data. Summary of the Invention

[0004] In order to solve the above technical problems, a big data analysis model evaluation method for market research and risk management is provided. This technical solution solves the problems raised in the above background technology.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] A big data analysis model evaluation method for market research and risk management, comprising:

[0007] Acquire at least one historical crisis event and construct a historical crisis event database;

[0008] Input historical crisis events into the big data analysis model for evaluation and obtain evaluation results;

[0009] Analyze the assessment results in conjunction with the historical crisis event database and obtain at least one assessment defect;

[0010] Evaluate the formation of failure points based on assessment defects;

[0011] Obtain important coefficients for evaluating fault points;

[0012] Based on the application scenario of the big data analysis model, the data used to evaluate the big data analysis model is predicted to obtain the data set to be tested;

[0013] Based on the assessed fault points, the data set to be tested is classified to obtain at least one data block to be tested;

[0014] Based on the assessed fault points, weights are assigned to the test data blocks;

[0015] Calculate the data evaluation perfection coefficient of the data block to be tested, accumulate it according to the weight of the data block to be tested, and obtain the total evaluation perfection coefficient of the big data analysis model.

[0016] Preferably, the acquiring of at least one historical crisis event and constructing a historical crisis event database comprises the following steps:

[0017] Obtain all analysis results of historical crisis events by existing data analysis models to obtain a set of analysis results;

[0018] The historical crisis events and analysis results are collected and summarized into a historical crisis event database.

[0019] Preferably, analyzing the evaluation results to obtain at least one evaluation defect comprises the following steps:

[0020] Compare the assessment results of historical crisis events with the set of historical crisis event analysis results to obtain the missing parts of the assessment;

[0021] Segment the missing part of the evaluation to obtain at least one missing local block, where the data types within the missing local block are consistent;

[0022] Obtain an existing data analysis model that is used to evaluate and generate missing local blocks as a target data analysis model;

[0023] In historical crisis events, the target data analysis model is used to generate data with missing local blocks as test data;

[0024] Monitor the analysis process of the big data analysis model to determine whether the test data is called. If so, use the path of the missing local block generated by the target data analysis model as the missing path of the big data analysis model. If not, use the test data as a blind spot for resource utilization.

[0025] The missing path is taken as the first evaluation defect, the resource utilization blind spot is taken as the second evaluation defect, and the first evaluation defect and the second evaluation defect are both taken as evaluation defects.

[0026] Preferably, the forming of the assessment fault point based on the assessment defect comprises the following steps:

[0027] Divide all resource usage blind areas into at least one blind area local block, where the data types in the blind area local blocks are consistent;

[0028] Perform blind spot data recognition on the local blind spot block to obtain at least one blind spot feature, and use the feature whose difference from the blind spot feature is less than a preset value as the blind spot approximate feature, where the preset value is the upper limit of the difference between two approximate features set based on experience;

[0029] Input the blind area approximate features into the big data analysis model. When the blind area approximate features cannot be identified, the blind area approximate features are used as blind area determination features. Otherwise, no processing is performed.

[0030] In historical crisis events, the data analyzed using the missing path are treated as data to be analyzed, and the data analyzed without the missing path are treated as data not to be analyzed;

[0031] Extract features from the data to be analyzed to obtain at least one feature to be analyzed, and use a feature whose difference from the feature to be analyzed is less than a preset value as an approximate feature to be analyzed;

[0032] Extract features from the data not to be analyzed to obtain at least one feature not to be analyzed, and use a feature whose difference from the feature not to be analyzed is less than a preset value as an approximate feature not to be analyzed;

[0033] Delete the features to be analyzed that are consistent with the features not to be analyzed, and use the deleted approximate features to be analyzed as missing trigger features;

[0034] Both blind spot determination features and missing trigger features are used as evaluation fault points.

[0035] Preferably, obtaining the important coefficients for evaluating the fault point includes the following steps:

[0036] Statistically assess the probability of failure points occurring in historical crisis events;

[0037] The historical crisis events with assessment failure points are used as target historical crisis events;

[0038] Obtain the missing analysis parts caused by the evaluation failure points in the big data analysis model;

[0039] Obtain the analysis results of the big data analysis model on the target historical crisis event as the target analysis result;

[0040] Obtaining a first response measure based on the target analysis result, and obtaining a second response measure based on the analysis missing portion and the target analysis result;

[0041] The difference between the second response and the first response is used as the characteristic coefficient for evaluating the fault point;

[0042] The characteristic coefficient is multiplied by the probability of occurrence of the assessed fault point to obtain the importance coefficient of the assessed fault point.

[0043] Preferably, the step of predicting the data evaluated by the big data analysis model to obtain a data set to be tested comprises the following steps:

[0044] Acquiring historical analysis data in at least one application scenario of the big data analysis model;

[0045] The amount of historical analysis data in the application scenario is used as the scale factor of the application scenario;

[0046] De-duplicate historical analysis data in the application scenario to obtain the data to be tested, and match the scale factor of the application scenario to the data to be tested;

[0047] The data to be tested and its corresponding proportional coefficients are summarized to obtain the data set to be tested.

[0048] Preferably, classifying the data set to be tested to obtain at least one data block to be tested comprises the following steps:

[0049] Extract features from the test data in the test data set to obtain at least one feature to be tested, where the feature to be tested is a feature whose difference from the assessed fault point is less than a preset value;

[0050] Count the number of features to be tested of a single test data as the feature value;

[0051] The proportional coefficient of the test data is divided by the characteristic value to obtain the assignment coefficient, and the assignment coefficient is matched to the test characteristic of the test data;

[0052] The features to be tested that have the smallest gap with the assessed fault point are classified into the same category to obtain the data blocks to be tested.

[0053] Preferably, assigning weights to the data blocks to be tested based on the assessed fault points comprises the following steps:

[0054] Accumulate the assignment coefficients of the features to be tested in the data block to be tested to obtain the preliminary weights;

[0055] The preliminary weight is multiplied by the importance coefficient of the assessed fault point corresponding to the data block to be tested to obtain the final weight of the data block to be tested.

[0056] Preferably, the step of calculating the data evaluation perfection coefficient of the data block to be tested comprises the following steps:

[0057] Obtain the average data output of all features to be tested in the data block to be tested in the existing data analysis model;

[0058] Obtain the data output of the data block to be tested in the big data analysis model;

[0059] The difference between the data output amount and the average data output amount is used to obtain the evaluation data amount;

[0060] The proportion of the evaluation data volume in the data output volume is used as the data evaluation perfection coefficient of the data block to be tested.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] By evaluating the formation of fault points based on evaluation defects, obtaining important coefficients of evaluation fault points, obtaining the data set to be tested and calculating the data evaluation perfection coefficient of the data block to be tested, the big data analysis model is evaluated from two aspects: data identification blind spots and missing data processing paths. In this way, the data analysis perfection of the big data analysis model in processing various types of data can be evaluated, and then the gap between it and the existing data analysis model can be compared, thereby more realistically reflecting the evaluation status of the big data analysis model. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 Schematic diagram of the process of the big data analysis model evaluation method for market research and risk management of the present invention;

[0064] Figure 2 A schematic diagram of a process for acquiring at least one historical crisis event and constructing a historical crisis event database according to the present invention;

[0065] Figure 3 A schematic diagram of a process for analyzing an evaluation result to obtain at least one evaluation defect according to the present invention;

[0066] Figure 4 This is a schematic diagram of the process of evaluating the formation of a fault point based on evaluation defects of the present invention;

[0067] Figure 5 A schematic diagram of a process for obtaining important coefficients for evaluating a fault point according to the present invention;

[0068] Figure 6 A schematic diagram of a process for predicting data for evaluating a big data analysis model of the present invention to obtain a data set to be tested;

[0069] Figure 7 A schematic diagram of a process of classifying a data set to be tested to obtain at least one data block to be tested according to the present invention;

[0070] Figure 8 This is a flow chart of assigning weights to test data blocks based on the assessed fault points of the present invention;

[0071] Figure 9 Schematic diagram of the flow of calculating the data evaluation perfection coefficient of the data block to be tested according to the present invention. DETAILED DESCRIPTION

[0072] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0073] Reference Figure 1 As shown, a big data analysis model evaluation method for market research and risk management includes:

[0074] Acquire at least one historical crisis event and construct a historical crisis event database;

[0075] Input historical crisis events into the big data analysis model for evaluation and obtain evaluation results;

[0076] Analyze the assessment results in conjunction with the historical crisis event database and obtain at least one assessment defect;

[0077] Evaluate the formation of failure points based on assessment defects;

[0078] Obtain important coefficients for evaluating fault points;

[0079] Based on the application scenario of the big data analysis model, the data used to evaluate the big data analysis model is predicted to obtain the data set to be tested;

[0080] Based on the assessed fault points, the data set to be tested is classified to obtain at least one data block to be tested;

[0081] Based on the assessed fault points, weights are assigned to the test data blocks;

[0082] Calculate the data evaluation perfection coefficient of the data block to be tested, accumulate it according to the weight of the data block to be tested, and obtain the total evaluation perfection coefficient of the big data analysis model.

[0083] In this solution, the big data analysis model is evaluated mainly from the perspective of data missing. During model analysis, data missing mainly focuses on two points. One is that there is a blind spot in the identification of the input data, that is, this part of the data is not processed. The other is that although the data is processed, the processing path is missing, resulting in incomplete analysis results. Either case will lead to the missing of analysis data, which in turn affects the results of the analysis. However, due to the uncertainty of the data used for analysis, the data missing situation is more complicated. At the same time, each type of missing has a different impact on completeness. Therefore, corresponding steps are set in the subsequent process to deal with the above-mentioned problems in a targeted manner.

[0084] Reference Figure 2 As shown, obtaining at least one historical crisis event and constructing a historical crisis event database includes the following steps:

[0085] Obtain all analysis results of historical crisis events by existing data analysis models to obtain a set of analysis results;

[0086] The historical crisis events and analysis results are collected and summarized into a historical crisis event database.

[0087] Reference Figure 3 As shown, analyzing the evaluation results to obtain at least one evaluation defect includes the following steps:

[0088] Compare the assessment results of historical crisis events with the set of historical crisis event analysis results to obtain the missing parts of the assessment;

[0089] Segment the missing part of the evaluation to obtain at least one missing local block, where the data types within the missing local block are consistent;

[0090] Obtain an existing data analysis model that is used to evaluate and generate missing local blocks as a target data analysis model;

[0091] In historical crisis events, the target data analysis model is used to generate data with missing local blocks as test data;

[0092] Monitor the analysis process of the big data analysis model to determine whether the test data is called. If so, use the path of the missing local block generated by the target data analysis model as the missing path of the big data analysis model. If not, use the test data as a blind spot for resource utilization.

[0093] The missing path is taken as the first evaluation defect, the resource utilization blind spot is taken as the second evaluation defect, and the first evaluation defect and the second evaluation defect are both taken as evaluation defects.

[0094] When identifying assessment defects, it is impossible to actively identify them through one's own assessment. A better way is to compare the analysis results with those of existing data analysis models. Therefore, historical crisis events are used as samples for analysis and comparison. Historical crisis events are specific and representative market research and risk management related events. Therefore, assessment defects can be obtained by comparing the analysis results.

[0095] Reference Figure 4 As shown, the formation of the assessment fault point based on the assessment defect includes the following steps:

[0096] Divide all resource usage blind areas into at least one blind area local block, where the data types in the blind area local blocks are consistent;

[0097] Perform blind spot data recognition on the local blind spot block to obtain at least one blind spot feature, and use the feature whose difference from the blind spot feature is less than a preset value as the blind spot approximate feature, where the preset value is the upper limit of the difference between two approximate features set based on experience;

[0098] Input the blind spot approximate features into the big data analysis model. When the blind spot approximate features cannot be identified, the blind spot approximate features are used as blind spot determination features. Otherwise, no processing is performed.

[0099] In historical crisis events, the data analyzed using the missing path are treated as data to be analyzed, and the data analyzed without the missing path are treated as data not to be analyzed;

[0100] Extract features from the data to be analyzed to obtain at least one feature to be analyzed, and use a feature whose difference from the feature to be analyzed is less than a preset value as an approximate feature to be analyzed;

[0101] Extract features from the data not to be analyzed to obtain at least one feature not to be analyzed, and use a feature whose difference from the feature not to be analyzed is less than a preset value as an approximate feature not to be analyzed;

[0102] Delete the features to be analyzed that are consistent with the features not to be analyzed, and use the deleted approximate features to be analyzed as missing trigger features;

[0103] Both blind spot determination features and missing trigger features are used as evaluation fault points.

[0104] Since the completeness of the analysis results of the data input into the big data analysis model needs to be analyzed later, it is necessary to identify the data features that cause missing analysis, and identify them from two aspects: identifying blind spots and path missing. When acquiring the features of path missing, it is necessary to compare the analysis of historical crisis events to obtain data for analysis of missing paths. As the data to be analyzed, the data to be analyzed here is the data to be analyzed for the missing path. Therefore, it will not be analyzed in the big data analysis model, which will cause the lack of analysis. Thus, the corresponding features are formed as missing trigger features. Here, due to the existence of approximate features, in order to ensure the completeness of feature recognition, it is necessary to acquire approximate features for the identified features and use the approximate features to acquire subsequent features, because these approximate features may also be required. When they are ignored, it will affect the accuracy of subsequent analysis.

[0105] Reference Figure 5 As shown in Figure 2, obtaining the important coefficients for evaluating the fault point includes the following steps:

[0106] Statistically assess the probability of failure points occurring in historical crisis events;

[0107] The historical crisis events with assessment failure points are used as target historical crisis events;

[0108] Obtain the missing analysis parts caused by the evaluation failure points in the big data analysis model;

[0109] Obtain the analysis results of the big data analysis model on the target historical crisis event as the target analysis result;

[0110] Obtaining a first response measure based on the target analysis result, and obtaining a second response measure based on the analysis missing portion and the target analysis result;

[0111] The difference between the second response and the first response is used as the characteristic coefficient for evaluating the fault point;

[0112] The characteristic coefficient is multiplied by the probability of occurrence of the assessed fault point to obtain the importance coefficient of the assessed fault point.

[0113] The data missing situations caused by the evaluation failure points are different because the amount of data missing caused by each evaluation failure point is different. Therefore, the decision impact caused by it is also different. In order to obtain its importance, it is necessary to determine it based on the change in its impact on the decision.

[0114] Reference Figure 6 As shown, predicting the data used to evaluate the big data analysis model to obtain the test data set includes the following steps:

[0115] Acquiring historical analysis data in at least one application scenario of the big data analysis model;

[0116] The amount of historical analysis data in the application scenario is used as the scale factor of the application scenario;

[0117] De-duplicate historical analysis data in the application scenario to obtain the data to be tested, and match the scale factor of the application scenario to the data to be tested;

[0118] The data to be tested and its corresponding proportional coefficients are summarized to obtain the data set to be tested.

[0119] The data that the big data analysis model analyzes is infinite. Therefore, in order to evaluate its analysis results, it is necessary to predict the distribution of the data to be analyzed. Thus, a test data set is generated. The test data set can, to a certain extent, predict the type distribution of the data that will be subsequently analyzed by the big data analysis model. Therefore, the completeness evaluation of the analysis results of the test data set can reflect the completeness of the data analysis of the big data analysis model.

[0120] Reference Figure 7 As shown, classifying the test data set to obtain at least one test data block includes the following steps:

[0121] Extract features from the test data in the test data set to obtain at least one feature to be tested, where the feature to be tested is a feature whose difference from the assessed fault point is less than a preset value;

[0122] Count the number of features to be tested of a single test data as the feature value;

[0123] The proportional coefficient of the test data is divided by the characteristic value to obtain the assignment coefficient, and the assignment coefficient is matched to the test characteristic of the test data;

[0124] The features to be tested that have the smallest gap with the assessed fault point are classified into the same category to obtain the data blocks to be tested.

[0125] In order to facilitate analysis, the data in the test data set is classified. The classification is based on the evaluation failure points, so that the same weight can be used to summarize the results of the test data blocks in the future.

[0126] Reference Figure 8 As shown, assigning weights to the test data blocks based on the assessed fault points includes the following steps:

[0127] Accumulate the assignment coefficients of the features to be tested in the data block to be tested to obtain the preliminary weights;

[0128] The preliminary weight is multiplied by the importance coefficient of the assessed fault point corresponding to the data block to be tested to obtain the final weight of the data block to be tested.

[0129] The data block to be tested contains different features to be tested, and the missing impact of each feature to be tested is different. Therefore, it is necessary to synthesize their impacts to obtain the weight of the data block to be tested.

[0130] Reference Figure 9 As shown, calculating the data evaluation perfection coefficient of the data block to be tested includes the following steps:

[0131] Obtain the average data output of all features to be tested in the data block to be tested in the existing data analysis model;

[0132] Obtain the data output of the data block to be tested in the big data analysis model;

[0133] The difference between the data output amount and the average data output amount is used to obtain the evaluation data amount;

[0134] The proportion of the evaluation data volume in the data output volume is used as the data evaluation perfection coefficient of the data block to be tested.

[0135] Furthermore, the present solution also proposes a storage medium on which a computer-readable program is stored. When the computer-readable program is called, the above-mentioned big data analysis model evaluation method for market research and risk management is executed.

[0136] It is understandable that the storage medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a DVD; or a semiconductor medium, such as a solid state disk (SSD).

[0137] To sum up, the advantages of the present invention are: by forming an evaluation fault point based on the evaluation defects, obtaining the important coefficients of the evaluation fault point, obtaining the data set to be tested and calculating the data evaluation perfection coefficient of the data block to be tested, the big data analysis model is evaluated from two aspects: data recognition blind spots and missing data processing paths, so that the data analysis perfection of the big data analysis model in processing various types of data can be evaluated, and then the gap between it and the existing data analysis model can be compared, so as to more realistically reflect the evaluation status of the big data analysis model.

[0138] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A big data analysis model evaluation method for market research and risk management, characterized by: include: Acquire at least one historical crisis event and construct a historical crisis event database; Input historical crisis events into the big data analysis model for evaluation and obtain evaluation results; Analyze the assessment results in conjunction with the historical crisis event database and obtain at least one assessment defect; Evaluate the formation of failure points based on assessment defects; Obtain important coefficients for evaluating fault points; Based on the application scenario of the big data analysis model, the data used to evaluate the big data analysis model is predicted to obtain the data set to be tested; Based on the assessed fault points, the data set to be tested is classified to obtain at least one data block to be tested; Based on the assessed fault points, weights are assigned to the test data blocks; Calculate the data evaluation perfection coefficient of the data block to be tested, accumulate it according to the weight of the data block to be tested, and obtain the total evaluation perfection coefficient of the big data analysis model.

2. A big data analysis model evaluation method for market research and risk management according to claim 1, characterized in that: Acquiring at least one historical crisis event and constructing a historical crisis event database comprises the following steps: Obtain all analysis results of historical crisis events by existing data analysis models to obtain a set of analysis results; The historical crisis events and analysis results are collected and summarized into a historical crisis event database.

3. A big data analysis model evaluation method for market research and risk management according to claim 2, characterized in that: Analyzing the evaluation results to obtain at least one evaluation defect comprises the following steps: Compare the assessment results of historical crisis events with the set of historical crisis event analysis results to obtain the missing parts of the assessment; Segment the missing part of the evaluation to obtain at least one missing local block, where the data types within the missing local block are consistent; Obtain an existing data analysis model that is used to evaluate and generate missing local blocks as a target data analysis model; In historical crisis events, the target data analysis model is used to generate data with missing local blocks as test data; Monitor the analysis process of the big data analysis model to determine whether the test data is called. If so, use the path of the missing local block generated by the target data analysis model as the missing path of the big data analysis model. If not, use the test data as a blind spot for resource utilization. The missing path is taken as the first evaluation defect, the resource utilization blind spot is taken as the second evaluation defect, and the first evaluation defect and the second evaluation defect are both taken as evaluation defects.

4. A big data analysis model evaluation method for market research and risk management according to claim 3, characterized in that: The formation of the assessment fault point based on the assessment defect comprises the following steps: Divide all resource usage blind areas into at least one blind area local block, where the data types in the blind area local blocks are consistent; Perform blind spot data recognition on the local blind spot block to obtain at least one blind spot feature, and use the feature whose difference from the blind spot feature is less than a preset value as the blind spot approximate feature, where the preset value is the upper limit of the difference between two approximate features set based on experience; Input the blind area approximate features into the big data analysis model. When the blind area approximate features cannot be identified, the blind area approximate features are used as blind area determination features. Otherwise, no processing is performed. In historical crisis events, the data analyzed using the missing path are treated as data to be analyzed, and the data analyzed without the missing path are treated as data not to be analyzed; Extract features from the data to be analyzed to obtain at least one feature to be analyzed, and use a feature whose difference from the feature to be analyzed is less than a preset value as an approximate feature to be analyzed; Extract features from the data not to be analyzed to obtain at least one feature not to be analyzed, and use a feature whose difference from the feature not to be analyzed is less than a preset value as an approximate feature not to be analyzed; Delete the features to be analyzed that are consistent with the features not to be analyzed, and use the deleted approximate features to be analyzed as missing trigger features; Both blind spot determination features and missing trigger features are used as evaluation fault points.

5. The big data analysis model evaluation method for market research and risk management according to claim 4, characterized in that: Obtaining the important coefficients for evaluating the fault point includes the following steps: Statistically assess the probability of failure points occurring in historical crisis events; The historical crisis events with assessment failure points are used as target historical crisis events; Obtain the missing analysis parts caused by the evaluation failure points in the big data analysis model; Obtain the analysis results of the big data analysis model on the target historical crisis event as the target analysis result; Obtaining a first response measure based on the target analysis result, and obtaining a second response measure based on the analysis missing portion and the target analysis result; The difference between the second response and the first response is used as the characteristic coefficient for evaluating the fault point; The characteristic coefficient is multiplied by the probability of occurrence of the assessed fault point to obtain the importance coefficient of the assessed fault point.

6. A big data analysis model evaluation method for market research and risk management according to claim 5, characterized in that: Predicting the data for evaluating the big data analysis model to obtain a test data set includes the following steps: Acquiring historical analysis data in at least one application scenario of the big data analysis model; The amount of historical analysis data in the application scenario is used as the scale factor of the application scenario; De-duplicate historical analysis data in the application scenario to obtain the data to be tested, and match the scale factor of the application scenario to the data to be tested; The data to be tested and its corresponding proportional coefficients are summarized to obtain the data set to be tested.

7. The big data analysis model evaluation method for market research and risk management according to claim 6, characterized in that: Classifying the data set to be tested to obtain at least one data block to be tested comprises the following steps: Extract features from the test data in the test data set to obtain at least one feature to be tested, where the feature to be tested is a feature whose difference from the assessed fault point is less than a preset value; Count the number of features to be tested of a single test data as the feature value; The proportional coefficient of the test data is divided by the characteristic value to obtain the assignment coefficient, and the assignment coefficient is matched to the test characteristic of the test data; The features to be tested that have the smallest gap with the assessed fault point are classified into the same category to obtain the data blocks to be tested.

8. The big data analysis model evaluation method for market research and risk management according to claim 7, characterized in that: The step of assigning weights to the test data blocks based on the assessed fault points includes the following steps: Accumulate the assignment coefficients of the features to be tested in the data block to be tested to obtain the preliminary weights; The preliminary weight is multiplied by the importance coefficient of the assessed fault point corresponding to the data block to be tested to obtain the final weight of the data block to be tested.

9. The big data analysis model evaluation method for market research and risk management according to claim 8, characterized in that: Calculating the data evaluation perfection coefficient of the data block to be tested includes the following steps: Obtain the average data output of all features to be tested in the data block to be tested in the existing data analysis model; Obtain the data output of the data block to be tested in the big data analysis model; The difference between the data output amount and the average data output amount is used to obtain the evaluation data amount; The proportion of the evaluation data volume in the data output volume is used as the data evaluation perfection coefficient of the data block to be tested.

Images