Financial service platform based on big data to complete predictive analysis

Abstract

The present invention relates to the field of predictive analysis technology, and in particular, to a financial service platform that performs predictive analysis based on big data. It includes an analysis data unit, a corresponding integration unit, a frequency judgment unit, and a similar integration unit. The corresponding integration unit of the present invention distributes data points to corresponding data groups, then obtains group data points from the data groups, and simultaneously integrates class data points in data types into type sets, calculates the data type center based on the integrated type set and class data points, and then calculates the distance from the group data points to the data type center, selects the data type center corresponding to the lowest class distance to find the data type, integrates the group data points into the corresponding data type, and then divides the data into a two-layer structure of data groups and data types, so that the integration process is clearer, the first step is to divide the data points into data groups, and the second step is to integrate the types according to the specific data groups, so that the hierarchical processing reduces the difficulty of integration and improves the accuracy of integration.

Description

Technical Field

[0001] The present invention relates to the field of predictive analysis technology, and in particular to a financial service platform that performs predictive analysis based on big data. Background Art

[0002] With the continuous evolution of big data technology, the application of predictive analysis will become more extensive and in-depth, becoming an important driving force for the financial services industry, and obtaining a large amount of financial data sources to help institutions understand customer needs, manage risks, and comply with compliance requirements. Since the large amount of financial data sources obtained contain diverse data, the data in the integration process is messy and scattered, which increases the difficulty of data integration and reduces the accuracy of integration. At the same time, when processing a large amount of diverse data, the data integration process contains useless calculations (data that is not used frequently). These calculations will consume computing resources (CPU, memory, storage), resulting in a decrease in the overall integration processing speed. Therefore, we provide a financial service platform based on big data to complete predictive analysis. Summary of the invention

[0003] The purpose of the present invention is to provide a financial service platform that performs predictive analysis based on big data to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above-mentioned purpose, the present invention provides a financial service platform for completing prediction analysis based on big data, including a data analysis unit, a corresponding integration unit, a frequency judgment unit, and a similarity integration unit;

[0005] The data analysis unit obtains a large amount of relevant financial data sources and historical data from the network to perform complete data analysis and predict integration difficulty. When the prediction difficulty is higher than a threshold, the analyzed complete data is regarded as different data points and the number of data points is recorded;

[0006] The corresponding integration unit is used to analyze the data in the data unit, calculate the data center of the data group according to the analyzed data in the data unit, and then calculate the distance from the data point to the data center according to the data center, the number of data groups and the data point, find the data group according to the data center corresponding to the distance, and allocate the data point to the corresponding data group;

[0007] The frequency judgment unit is used to receive the data group in the corresponding integration unit, calculate the characteristic value of the group data point according to the group data points and the number of group data points in the data group, and judge the exclusion status of the group data point according to the calculated characteristic value of the group data point;

[0008] The corresponding integration unit receives the group data points in the frequency judgment unit and the data types in the analysis data unit, and records the number of class data points and the number of data types, calculates the data type center according to the class data points and then calculates the distance from the group data points to the data type center, then finds the data type according to the data type center corresponding to the distance, and integrates the group data points into the corresponding data type;

[0009] The similarity integration unit is used to receive commands with low integration difficulty in the analysis data unit. The similarity integration unit obtains the analyzed complete data source and type feature keywords from the analysis data unit to calculate the similarity, and integrates the data in the analyzed complete data source into the corresponding data type according to the calculated similarity.

[0010] As a further improvement of the technical solution, the data analysis unit includes an analysis prediction module and a data set module;

[0011] The analysis and prediction module uses crawler technology to capture a large number of relevant financial data sources on the Internet and obtain historical data. It uses machine learning in big data to perform complete data analysis on the captured large number of relevant financial data sources, uses neural networks in big data to predict the integration difficulty of the analyzed complete data sources, and then uses the predicted integration difficulty and the standard integration difficulty to judge the difficulty. The specific judgment situations include:

[0012] Case ①: When the predicted integration difficulty is greater than the standard integration difficulty, it means that the integration difficulty is high when processing the complete data source for analysis, and the command with high integration difficulty is passed to the data set module;

[0013] Case ②: When the predicted integration difficulty is less than the standard integration difficulty, it means that the integration difficulty is low when processing the complete data source for analysis, and the command with low integration difficulty is transferred to the similar integration unit;

[0014] Historical data includes data groups and data types;

[0015] Data group market data group, customer data group, transaction data group, number of data groups;

[0016] The data types include time series type (including type feature keywords stock price, economic indicators), financial statement type (including type feature keywords balance sheet, income statement, cash flow statement), text type (including type feature keywords news report, company announcement, market comment), and data type quantity;

[0017] The data set module is used to receive commands with high integration difficulty in the analysis and prediction module. The data set module obtains the complete data source and data group for analysis from the analysis and prediction module. Combine the complete data sources for analysis into a data set, treat the data in the combined data set as different data points, and record the number of data points;

[0018] As a further improvement of the present technical solution, the corresponding integration unit is used to receive the number of data points, data points, integrated data sets in the data set module and the number of data groups, data groups in the analysis and prediction module, calculate the data center of the data group according to the integrated data set and data points, and then calculate the distance from the data point to the data center according to the data center, the number of data groups, the integrated data set and data points, and so on, to obtain the distance from the data point to different data centers, and then compare the distances from the data point to different data centers in turn, select the lowest distance, find the data group according to the data center corresponding to the lowest distance, and allocate the data point to the corresponding data group. By calculating the distance from the data point to each data center, it can be effectively determined to which data group each data point should be allocated. This distance-based allocation method can improve the accuracy of classification, which is convenient for the subsequent group of data points to be accurately integrated into the corresponding data type;

[0019] The data center implementation principle of the calculation data group:

[0020] Collecting integrated datasets and data points Calculate the data center of the data group and get the data center , the specific algorithm formula is:

[0021] ;

[0022] in, Refers to the Data centers;

[0023] Principle of calculating the distance from the data center to the data group:

[0024] Collection Data Center , Number of data sets , integrated dataset and data points Calculate the distance from the data center to the data group and get the distance , the specific algorithm formula is:

[0025] ;

[0026] As a further improvement of the technical solution, the frequency judgment unit is used to receive the data group in the corresponding integration unit, take the group data points and the number of group data points from the data group, calculate the mean of the group data points according to the group data points and the number of group data points in the data group, and then calculate the standard deviation of the group data points according to the group data points, the number of group data points and the mean of the group data points, calculate the characteristic value of the group data points using the group data points, the number of group data points, the mean of the group data points and the standard deviation of the group data points, judge the frequency of feature utilization according to the calculated characteristic value of the group data points and the standard group characteristic value, and judge the retention and exclusion of data points by using the frequency to reduce useless calculation amount;

[0027] Principle of calculating the mean of group data points:

[0028] Collect data points according to group and the number of group data points Calculate the mean of the group data points and get the mean of the group data points , the specific algorithm formula is:

[0029] ;

[0030] in, Refers to the first The formula is used to calculate the mean of the group data points, which is convenient for the subsequent calculation of the characteristic values ​​of the group data points;

[0031] The implementation principle of calculating the standard deviation of group data points:

[0032] Collecting Group Data Points , number of group data points and the mean of the group data points Calculate the standard deviation of the group data points and get the standard deviation of the group data points , the specific algorithm formula is:

[0033] ;

[0034] Among them, this formula is used to calculate the standard deviation of the group data points, which is convenient for the subsequent calculation of the characteristic values ​​of the group data points.

[0035] As a further improvement of the technical solution, the implementation principle of calculating the characteristic value of the group data point in the frequency judgment unit is:

[0036] Group Data Points and the number of group data points , group data point mean and group data point standard deviation Calculate the eigenvalues ​​of the group data points and obtain the eigenvalues ​​of the group data points , the specific algorithm formula is:

[0037] ;

[0038] in, The fourth power in represents the fourth power of the deviation of each group data point from the mean, which is to emphasize the utilization frequency of the group data point. The characteristic value of the group data point is standardized. Since the standardized characteristic value of the group data point is 0, the purpose of subtracting 3 is to adjust the characteristic value of the group data point to a value comparable to the standard group characteristic value;

[0039] As a further improvement of the technical solution, the frequency judgment unit specifically judges the following:

[0040] Case 1: When the calculated group data point eigenvalue is greater than or equal to the standard group eigenvalue, it means that the group data point is used more frequently, and the group data point is fed back to the corresponding integration unit;

[0041] Case 2: When the calculated eigenvalue of a group data point is less than the standard group eigenvalue, it means that the utilization frequency of the group data point is low, and the group data point is directly excluded.

[0042] As a further improvement of the present technical solution, the corresponding integration unit receives the group data points in the frequency judgment unit and obtains the data type in the analysis and prediction module, integrates the class data points in each data type into a type set, and records the number of class data points and the number of data types, and calculates the data type center according to the integrated type set and class data points (the implementation principle of calculating the data type center is the same as the implementation principle of calculating the data center of the above-mentioned data group, and the integrated data set and data points are replaced with the integrated type set and class data points), and then calculates the distance from the group data point to the data type center according to the data type center, the number of data types, the integrated type set and the group data points (the implementation principle of calculating the distance from the group data point to the data type center is the same as The principle and operation of calculating the distance from the data center to the data group are the same. The data center, the number of data groups, the integrated data set and the data point are replaced with the data type center, the number of data types, the integrated type set and the group data point, and so on. The distance from the group data point to the center of different data types is obtained, and then the distances from the group data point to the centers of different data types are compared in turn. The lowest class distance is selected, and the data type is found according to the data type center corresponding to the lowest class distance. The group data points are integrated into the corresponding data type. Data integration is performed according to the lowest class distance, which can make each data point consistent with the feature vector of its category. This method makes the data integration process more orderly and improves the consistency of the integrated data.

[0043] As a further improvement of the technical solution, the similar integration unit is used to receive commands with low integration difficulty in the analysis and prediction module, the similar integration unit obtains the complete data source and type feature keywords of the analysis from the analysis and prediction module, uses the feature extraction technology to extract the feature keywords of each data in the complete data source of the analysis, and uses The vector method calculates the feature vectors of the extracted feature keywords and type feature keywords, and then calculates the similarity between the extracted feature vectors and feature vectors of different types. According to the calculated similarity, the different data in the analyzed complete data source are integrated into the corresponding data type. After the data is integrated, if new data needs to be added or updated, it only needs to be operated according to the established data type, which reduces repetitive work, especially in data migration and conversion, and greatly improves work efficiency. The specific integration situation is as follows:

[0044] Case 1: When the calculated similarity is close to When , it means that the extracted feature keyword is similar to the type feature keyword, and then the similarities are calculated and compared in sequence, the highest similarity is selected, and the data corresponding to the extracted feature keyword is integrated into the data type corresponding to the highest similarity;

[0045] Case 2: When the calculated similarity is close to When , it means that the extracted feature keywords and the type feature keywords are not similar, and the data corresponding to the extracted feature keywords are directly excluded;

[0046] The implementation steps for calculating similarity are:

[0047] Step 1: Collect the extracted feature vectors and type feature vector , according to the extracted feature vector Calculate the length of the extracted feature vector , the specific algorithm formula is:

[0048] ;

[0049] Among them, this formula is used to calculate the length of the extracted feature vector, which is convenient for subsequent similarity calculation; Step ②, according to the type feature vector Compute the length of a type's characteristic vector , the specific algorithm formula is:

[0050] ;

[0051] Among them, this formula is used to calculate the length of the type feature vector, which is convenient for subsequent calculation of similarity;

[0052] Step 3: Use the extracted feature vector , type feature vector , extract the length of the feature vector and the length of the type character vector Calculate the similarity and get the calculated similarity , the specific algorithm formula is:

[0053] ;

[0054] in, The value range is This formula is used to calculate the similarity. By calculating the similarity between feature vectors, similar data in the data source can be integrated more accurately, and the integration errors can be reduced, thus improving the accuracy of data integration.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. In the financial service platform that completes predictive analysis of big data, the corresponding integration unit calculates the data center of the data group based on the integrated data set and data points, and then calculates the distance from the data point to the data center based on the data center, selects the data center corresponding to the lowest distance to find the data group, assigns the data points to the corresponding data group, and then obtains the group data points from the data group. At the same time, the class data points in each data type are integrated into a type set, and the data type center is calculated based on the integrated type set and class data points, and then the distance from the group data point to the data type center is calculated, the data type is found by selecting the data type center corresponding to the lowest class distance, and then the group data points are integrated into the corresponding data type. The data is divided into a two-layer structure of data group and data type, making the integration process clearer. The first step is to divide the data points into related data groups, and the second step is to integrate the data group according to the specific integration type, so that the hierarchical processing reduces the difficulty of integration, facilitates understanding and implementation, and improves the accuracy of integration.

[0057] 2. In the financial service platform that completes predictive analysis based on big data, the frequency judgment unit takes group data points and the number of group data points from the data group, calculates the mean of the group data points based on the group data points and the number of group data points in the data group, and then calculates the standard deviation of the group data points based on the group data points, the number of group data points and the mean of the group data points. The group data points are calculated using the group data points, the number of group data points, the mean of the group data points and the standard deviation of the group data points. The frequency of feature utilization is judged based on the calculated group data point characteristic values ​​and the standard group characteristic values. The frequency is used to judge whether to retain or exclude data points, thereby reducing the required time and useless calculations and improving the processing speed of integration. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is an overall block diagram of the present invention;

[0059] Figure 2 It is a block diagram of the data analysis unit of the present invention.

[0060] The meaning of each number in the figure is:

[0061] 1. Analysis data unit; 11. Analysis and prediction module; 12. Data set module; 2. Corresponding integration unit;

[0062] 3. Frequency judgment unit; 4. Similarity integration unit. DETAILED DESCRIPTION

[0063] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] The present invention provides a financial service platform based on big data to complete predictive analysis, please refer to Figure 1-Figure 2 , including a data analysis unit 1, a corresponding integration unit 2, a frequency judgment unit 3, and a similarity integration unit 4;

[0065] The data analysis unit 1 obtains a large amount of relevant financial data sources and historical data from the network to perform complete data analysis and predict the integration difficulty. When the prediction difficulty is higher than the threshold, the analyzed complete data is regarded as different data points and the number of data points is recorded; the corresponding integration unit 2 is used to analyze the data in the data unit 1, calculate the data center of the data group based on the data in the analysis data unit 1, and then calculate the distance from the data point to the data center based on the data center, the number of data groups and the data points, find the data group according to the data center corresponding to the distance, and assign the data point to the corresponding data group; the frequency judgment unit 3 is used to receive the data group in the corresponding integration unit 2, calculate the characteristic value of the group data point according to the group data points and the number of group data points in the data group, and calculate the characteristic value of the group data point according to the calculated The characteristic value of the group data point determines the exclusion of the group data point; the corresponding integration unit 2 receives the group data point in the frequency judgment unit 3 and the data type in the analysis data unit 1, and records the number of class data points and the number of data types, calculates the data type center according to the class data point and then calculates the distance from the group data point to the data type center, and then finds the data type according to the data type center corresponding to the distance, and integrates the group data point into the corresponding data type; the similarity integration unit 4 is used to receive the command with low integration difficulty in the analysis data unit 1, the similarity integration unit 4 obtains the analyzed complete data source and type feature keywords from the analysis data unit 1 to calculate the similarity, and integrates the data in the analyzed complete data source into the corresponding data type according to the calculated similarity.

[0066] The following is a refinement of the above units, see Figure 1-Figure 2 ;

[0067] The data analysis unit 1 includes an analysis and prediction module 11 and a data set module 12;

[0068] The analysis and prediction module 11 uses crawler technology to capture a large number of relevant financial data sources on the Internet and obtain historical data. It uses machine learning in big data to perform complete data analysis on the captured large number of relevant financial data sources, uses the neural network in big data to predict the integration difficulty of the analyzed complete data source, and then uses the predicted integration difficulty and the standard integration difficulty to make a difficulty judgment. The specific judgment situations include:

[0069] Case ①: when the predicted integration difficulty is greater than the standard integration difficulty, it means that the integration difficulty is high when processing the complete data source for analysis, and the command with high integration difficulty is transmitted to the data set module 12;

[0070] Case ②, when the predicted integration difficulty is less than the standard integration difficulty, it means that when processing the complete data source for analysis, the integration difficulty is low, and the command with low integration difficulty is transmitted to the similar integration unit 4;

[0071] Historical data includes data groups and data types;

[0072] Data group market data group, customer data group, transaction data group, number of data groups;

[0073] The data types include time series type (including type feature keywords stock price, economic indicators), financial statement type (including type feature keywords balance sheet, income statement, cash flow statement), text type (including type feature keywords news report, company announcement, market comment), and data type quantity;

[0074] The data set module 12 is used to receive commands with high integration difficulty from the analysis and prediction module 11. The data set module 12 obtains the complete data source and data group for analysis from the analysis and prediction module 11. Combine the complete data sources for analysis into a data set, treat the data in the combined data set as different data points, and record the number of data points;

[0075] The corresponding integration unit 2 is used to receive the number of data points, data points, integrated data sets in the data set module 12 and the number of data groups and data groups in the analysis and prediction module 11, calculate the data center of the data group according to the integrated data set and data points, and then calculate the distance from the data point to the data center according to the data center, the number of data groups, the integrated data set and data points, and so on, to obtain the distance from the data point to different data centers, and then compare the distances from the data points to different data centers in turn, select the lowest distance, find the data group according to the data center corresponding to the lowest distance, and allocate the data points to the corresponding data groups. By calculating the distance from the data point to each data center, it can be effectively determined to which data group each data point should be allocated. This distance-based allocation method can improve the accuracy of classification, which is convenient for the subsequent group of data points to be accurately integrated into the corresponding data type;

[0076] The data center implementation principle of the calculation data group:

[0077] Collecting integrated datasets and data points Calculate the data center of the data group and get the data center , the specific algorithm formula is:

[0078] ;

[0079] in, Refers to the Data centers;

[0080] Principle of calculating the distance from the data center to the data group:

[0081] Collection Data Center , Number of data sets , integrated dataset and data points Calculate the distance from the data center to the data group and get the distance , the specific algorithm formula is:

[0082] ;

[0083] The frequency judgment unit 3 is used to receive the data group in the corresponding integration unit 2, take the group data points and the number of group data points from the data group, calculate the mean of the group data points according to the group data points and the number of group data points in the data group, and then calculate the standard deviation of the group data points according to the group data points, the number of group data points and the mean of the group data points, calculate the characteristic value of the group data points by using the group data points, the number of group data points, the mean of the group data points and the standard deviation of the group data points, judge the frequency of feature utilization according to the calculated characteristic value of the group data points and the standard group characteristic value, judge the retention and exclusion of data points by using the frequency, and reduce the amount of useless calculation;

[0084] Principle of calculating the mean of group data points:

[0085] Collect data points according to group and the number of group data points Calculate the mean of the group data points and get the mean of the group data points , the specific algorithm formula is:

[0086] ;

[0087] in, Refers to the first The formula is used to calculate the mean of the group data points, which is convenient for the subsequent calculation of the characteristic values ​​of the group data points;

[0088] The implementation principle of calculating the standard deviation of group data points:

[0089] Collecting Group Data Points , number of group data points and the mean of the group data points Calculate the standard deviation of the group data points and get the standard deviation of the group data points , the specific algorithm formula is:

[0090] ;

[0091] Among them, this formula is used to calculate the standard deviation of the group data points, which is convenient for the subsequent calculation of the characteristic values ​​of the group data points;

[0092] The implementation principle of calculating the characteristic value of the group data point in the frequency judgment unit 3 is:

[0093] Group Data Points and the number of group data points , group data point mean and group data point standard deviation Calculate the eigenvalues ​​of the group data points and obtain the eigenvalues ​​of the group data points , the specific algorithm formula is:

[0094] ;

[0095] in, The fourth power in represents the fourth power of the deviation of each group data point from the mean, which is to emphasize the utilization frequency of the group data point. The characteristic value of the group data point is standardized. Since the standardized characteristic value of the group data point is 0, the purpose of subtracting 3 is to adjust the characteristic value of the group data point to a value comparable to the standard group characteristic value;

[0096] Specific judgment situation in frequency judgment unit 3:

[0097] Case 1: When the calculated group data point characteristic value is greater than or equal to the standard group characteristic value, it means that the group data point is used more frequently, and the group data point is fed back to the corresponding integration unit 2;

[0098] Case 2: When the calculated group data point eigenvalue is less than the standard group eigenvalue, it means that the group data point is less frequently used, and the group data point is directly excluded;

[0099] The corresponding integration unit 2 receives the group data points in the frequency judgment unit 3 and obtains the data type in the analysis and prediction module 11, integrates the class data points in each data type into a type set, and records the number of class data points and the number of data types, and calculates the data type center according to the integrated type set and class data points (the implementation principle of calculating the data type center is the same as the implementation principle of calculating the data center of the data group mentioned above, replacing the integrated data set and data points with the integrated type set and class data points), and then calculates the distance from the group data point to the data type center according to the data type center, the number of data types, the integrated type set and the group data points (the implementation principle of calculating the distance from the group data point to the data type center is the same as the implementation principle of calculating the data center The distance to the data group is realized in the same principle. The data center, the number of data groups, the integrated data set and the data point are replaced with the data type center, the number of data types, the integrated type set and the group data point. By analogy, the distance from the group data point to the center of different data types is obtained. Then the distance from the group data point to the center of different data types is compared in turn. The lowest class distance is selected. The data type is found according to the data type center corresponding to the lowest class distance. The group data points are integrated into the corresponding data type. Data integration is performed according to the lowest class distance, which can make each data point consistent with the feature vector of its category. This method makes the data integration process more orderly and improves the consistency of the integrated data.

[0100] The similar integration unit 4 is used to receive the command with low integration difficulty in the analysis and prediction module 11, obtain the complete data source and type feature keywords of the analysis from the analysis and prediction module 11, extract the feature keywords of each data in the complete data source of the analysis by using the feature extraction technology, and use The vector method calculates the feature vectors of the extracted feature keywords and type feature keywords, and then calculates the similarity between the extracted feature vectors and feature vectors of different types. According to the calculated similarity, the different data in the analyzed complete data source are integrated into the corresponding data type. After the data is integrated, if new data needs to be added or updated, it only needs to be operated according to the established data type, which reduces repetitive work, especially in data migration and conversion, and greatly improves work efficiency. The specific integration situation is as follows:

[0101] Case 1: When the calculated similarity is close to When , it means that the extracted feature keyword is similar to the type feature keyword, and then the similarities are calculated and compared in sequence, the highest similarity is selected, and the data corresponding to the extracted feature keyword is integrated into the data type corresponding to the highest similarity;

[0102] Case 2: When the calculated similarity is close to When , it means that the extracted feature keywords and the type feature keywords are not similar, and the data corresponding to the extracted feature keywords are directly excluded;

[0103] The implementation steps for calculating similarity are:

[0104] Step 1: Collect the extracted feature vectors and type feature vector , according to the extracted feature vector Calculate the length of the extracted feature vector , the specific algorithm formula is:

[0105] ;

[0106] Among them, this formula is used to calculate the length of the extracted feature vector, which is convenient for subsequent similarity calculation; Step ②, according to the type feature vector Compute the length of a type's characteristic vector , the specific algorithm formula is:

[0107] ;

[0108] Among them, this formula is used to calculate the length of the type feature vector, which is convenient for subsequent calculation of similarity;

[0109] Step 3: Use the extracted feature vector , type feature vector , extract the length of the feature vector and the length of the type character vector Calculate the similarity and get the calculated similarity , the specific algorithm formula is:

[0110] ;

[0111] in, The value range is This formula is used to calculate the similarity. By calculating the similarity between feature vectors, similar data in the data source can be integrated more accurately, and the integration errors can be reduced, thus improving the accuracy of data integration.

[0112] Usage process:

[0113] The corresponding integration unit 2 is used to receive the data in the data set module 12 and the analysis and prediction module 11, and then calculate the distance from the data point to the data center according to the data center, the number of data groups, the integrated data set and the data point in the data set module 12 and the analysis and prediction module 11, and obtain the distance from the data point to different data centers, and then compare the distances from the data point to different data centers in turn, select the data center corresponding to the lowest distance, and find the data group, assign the data point to the corresponding data group, and then pass the data group to the frequency judgment unit 3, the frequency judgment unit 3 takes the group data points and the number of group data points from the data group, calculates the mean of the group data points, and then calculates the standard deviation of the group data points according to the group data points, the number of group data points and the mean of the group data points, and uses the group data points, the number of group data points, the mean of the group data points and the standard deviation of the group data points to calculate the characteristic value of the group data points , the feature utilization frequency is judged according to the calculated group data point characteristic value and the standard group characteristic value. When the calculated group data point characteristic value is greater than or equal to the standard group characteristic value, it means that the group data point utilization frequency is high, and the group data point is fed back to the corresponding integration unit 2. The corresponding integration unit 2 integrates the class data points in each data type into a type set, and records the number of class data points and the number of data types. The data type center is calculated according to the integrated type set and the class data point, and then the distance from the group data point to the data type center is calculated according to the data type center, the number of data types, the integrated type set and the group data point, and the distance from the group data point to the center of the data type is obtained. The distance from the group data point to the center of different data types is then compared in turn, and the data type center corresponding to the lowest class distance is selected and the data type is found, and the group data point is integrated into the corresponding data type.

[0114] 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 by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A financial service platform that performs predictive analysis based on big data, characterized by: It includes a data analysis unit (1), a corresponding integration unit (2), a frequency judgment unit (3), and a similarity integration unit (4); The data analysis unit (1) obtains a large amount of relevant financial data sources and historical data from the network to perform complete data analysis and predict integration difficulty. When the prediction difficulty is higher than a threshold, the analyzed complete data is regarded as different data points and the number of data points is recorded; The corresponding integration unit (2) is used to analyze the data in the data unit (1), calculate the data center of the data group according to the analyzed data in the data unit (1), and then calculate the distance from the data point to the data center according to the data center, the number of data groups and the data points, find the data group according to the data center corresponding to the distance, and allocate the data point to the corresponding data group; The frequency judgment unit (3) is used to receive the data group in the corresponding integration unit (2), calculate the characteristic value of the group data point according to the group data points and the number of group data points in the data group, and judge the exclusion status of the group data point according to the calculated characteristic value of the group data point; The corresponding integration unit (2) receives the group data points in the frequency judgment unit (3) and the data types in the analysis data unit (1), and records the number of class data points and the number of data types, calculates the data type center according to the class data points and then calculates the distance from the group data points to the data type center, then finds the data type according to the data type center corresponding to the distance, and integrates the group data points into the corresponding data type; The similarity integration unit (4) is used to receive commands with low integration difficulty from the analysis data unit (1), the similarity integration unit (4) obtains data from the analysis data unit (1) to calculate similarity, and integrates the data in the analyzed complete data source into the corresponding data type according to the calculated similarity.

2. The financial service platform for predictive analysis based on big data according to claim 1, characterized in that: The data analysis unit (1) comprises an analysis prediction module (11) and a data set module (12); The analysis and prediction module (11) uses crawler technology to capture a large number of relevant financial data sources on the Internet and obtain historical data. It uses machine learning in big data to perform complete data analysis on the captured large number of relevant financial data sources, uses neural networks in big data to predict the integration difficulty of the analyzed complete data sources, and then uses the predicted integration difficulty and the standard integration difficulty to make a difficulty judgment. The specific judgment includes: Case ①: When the predicted integration difficulty is greater than the standard integration difficulty, it means that the integration difficulty is high when processing the complete data source for analysis, and the command with high integration difficulty is passed to the data set module (12); Case ②: When the predicted integration difficulty is less than the standard integration difficulty, it means that the integration difficulty is low when processing the complete data source for analysis, and the command with low integration difficulty is transmitted to the similar integration unit (4); The data set module (12) is used to receive commands with high integration difficulty in the analysis and prediction module (11), and the data set module (12) obtains the complete data source and data group for analysis from the analysis and prediction module (11), and uses The complete data sources for analysis are integrated into a data set, the data in the integrated data set are treated as different data points, and the number of data points is recorded.

3. The financial service platform for predictive analysis based on big data according to claim 2, characterized in that: The corresponding integration unit (2) is used to receive the number of data points, data points, and integrated data sets in the data set module (12) and the number of data groups and data groups in the analysis and prediction module (11), calculate the data center of the data group based on the integrated data set and data points, and then calculate the distance from the data point to the data center based on the data center, the number of data groups, the integrated data set and data points, and then compare the distances from the data point to different data centers in turn, select the lowest distance, find the data group based on the data center corresponding to the lowest distance, and then allocate the data point to the corresponding data group.

4. The financial service platform for predictive analysis based on big data according to claim 3, characterized in that: The frequency judgment unit (3) is used to receive the data group in the corresponding integration unit (2), take the group data points and the number of group data points from the data group, calculate the mean of the group data points according to the group data points and the number of group data points in the data group, and then calculate the standard deviation of the group data points according to the group data points, the number of group data points and the mean of the group data points, calculate the characteristic value of the group data points using the group data points, the number of group data points, the mean of the group data points and the standard deviation of the group data points, and judge the frequency of feature utilization according to the calculated characteristic value of the group data points and the standard group characteristic value.

5. The financial service platform for predictive analysis based on big data according to claim 4, characterized in that: The implementation principle of calculating the characteristic value of the group data point in the frequency judgment unit (3) is: Group Data Points and the number of group data points , group data point mean and group data point standard deviation Calculate the eigenvalues ​​of the group data points and obtain the eigenvalues ​​of the group data points , the specific algorithm formula is: ； in, The fourth power in represents the fourth power of the deviation of each group of data points from the mean.

6. The financial service platform for predictive analysis based on big data according to claim 4, characterized in that: The specific judgment situation in the frequency judgment unit (3) is: Case 1: When the calculated group data point characteristic value is greater than or equal to the standard group characteristic value, it means that the group data point is used more frequently, and the group data point is fed back to the corresponding integration unit (2); Case 2: When the calculated eigenvalue of a group data point is less than the standard group eigenvalue, it means that the utilization frequency of the group data point is low, and the group data point is directly excluded.

7. The financial service platform for predictive analysis based on big data according to claim 4, characterized in that: The corresponding integration unit (2) receives the group data points in the frequency judgment unit (3) and the data type in the analysis and prediction module (11), integrates the class data points in the data type into a type set, and records the number of class data points and the number of data types, calculates the data type center based on the integrated type set and the class data points, and then calculates the distance from the group data point to the data type center based on the data type center, compares the distances from the group data point to the centers of different data types in turn, selects the lowest class distance, finds the data type based on the data type center corresponding to the lowest class distance, and integrates the group data point into the corresponding data type.

8. The financial service platform for predictive analysis based on big data according to claim 7, characterized in that: The similarity integration unit (4) is used to receive commands with low integration difficulty in the analysis and prediction module (11), the similarity integration unit (4) obtains the complete data source and type feature keywords for analysis from the analysis and prediction module (11), uses feature extraction technology to extract feature keywords of each data in the complete data source for analysis, and uses The vector method calculates the feature vectors of the extracted feature keywords and type feature keywords, and then calculates the similarity between the extracted feature vectors and different types of feature vectors. According to the calculated similarity, the different data in the analyzed complete data source are integrated into the corresponding data type.

Images