A big data testing system and method based on matrix logic

Abstract

The application discloses a big data testing system and method based on matrix logic, which comprises a big data acquisition module, a big data processing module, a big data testing module and a big data storage module, and the modules are connected in communication with each other; three testing stages are carried out in the ETL process of big data, including data stage verification testing, MapReduce stage testing and output data stage verification testing, and the data of big data testing is synchronized to the big data storage module; the third stage output stage verification of big data testing is tested, the accuracy, uniqueness, integrity, consistency, correlation and timeliness of data are tested, the quality of data is ensured, and the data is better and more accurate in application to decision analysis and recommendation system on the business.

Description

Technical Field

[0001] This invention belongs to the field of big data testing, and in particular, it applies the matrix logic concept to complete the testing task after classification using the distributed database HBase in the third stage of big data testing. Specifically, it involves a big data testing system and method based on the matrix logic concept. Background Technology

[0002] Big data has become synonymous with an era. The arrival of the big data era has overturned the conventional way of thinking about data. With the rise of mobile internet and smart devices, more and more data is being accumulated on the application platforms of various companies. This data, which contains a large number of user characteristics and behavior logs, is stored in massive quantities. After statistical analysis and feature sample extraction, and then training, corresponding business algorithm models are produced. It can accurately identify and predict user behavior and intentions. However, before using it in the above-mentioned decision analysis or recommendation systems, how to ensure the accuracy and compliance of the data, and the testing of big data, are particularly important. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a big data testing system and method based on matrix logic, which can overcome the shortcomings of the prior art.

[0004] The technical solution of this invention is:

[0005] A big data testing system based on matrix logic includes a big data acquisition module, a big data processing module, a big data testing module, and a big data storage module, with each module communicating and connecting with the others.

[0006] The big data acquisition module is used to collect data from the data source;

[0007] The big data processing module is used to process the ETL process of the data;

[0008] The big data testing module is used to test the ETL process of the data;

[0009] The big data storage module is used to synchronize the data after the big data test is completed and the data has been modified and repaired to the target data warehouse in HIVE.

[0010] The aforementioned data sources include sensors, smart devices, online systems, offline systems, social networks, and internet platforms; the data acquisition methods include database acquisition, system log acquisition, network data acquisition, and sensor device data acquisition.

[0011] The aforementioned big data processing module includes online processing of data extraction, transformation, and loading, as well as data cleaning, integration, transformation, and reduction during MapReduce data processing.

[0012] The aforementioned big data testing module includes three testing phases: Phase 1: Data phase verification test; Phase 2: MapReduce phase test; Phase 3: Output data phase verification test, and the data after the big data test is completed and the changes and repairs are synchronized to the big data storage module.

[0013] A big data testing method based on matrix logic is characterized by the following steps:

[0014] S1. During the data ETL process, conduct the first stage of big data testing: data phase testing and verification.

[0015] S2. During the data ETL process, conduct the second phase of big data testing: MapReduce phase testing and verification;

[0016] S3. In the data ETL process, the third stage of big data testing is carried out: output data stage verification test. After the above MapReduce stage test verification, the results of the MapReduce process are obtained, and the result data is written to the distributed file system HDFS. Based on HDFS as the underlying support, these data results are used to generate the high-concurrency distributed database HBase.

[0017] S4. In distributed HBase, classification is based on its data model-enhanced coefficient sorting mapping table, including row key, column key, or timestamp.

[0018] S5. In the matrix frame model, create a matrix frame based on the number of categories. The horizontal input of the matrix frame is the test cases of the data to be tested.

[0019] In the aforementioned big data testing method, step S1 includes the following testing and verification contents: testing and verifying whether the data resources of various aspects are correctly loaded into the Hadoop system, testing and verifying whether the source data matches the data pushed into the Hadoop system, and testing and verifying whether the correct data is extracted and loaded into the correct location.

[0020] In the aforementioned big data testing method, step S2 includes the following second-stage testing and verification content: during the MapReduce processing of data in HDFS, in the entire distributed device system, testing and verifying whether MapReduce runs normally on a single node and multiple nodes, testing and verifying whether the output file is correct, testing and verifying whether the key-value pairs in the MapReduce processing are correctly generated, and testing and verifying whether the data aggregation and merging is correct after the Reduce process is completed.

[0021] In the aforementioned big data testing method, step S5 includes test cases for the data to be tested, such as accuracy, uniqueness, completeness, consistency, relevance, and timeliness.

[0022] In the aforementioned big data testing method, step S5 involves using the AdaBoost iterative algorithm horizontally to create weak classifiers for the test cases of the current data to be tested; and using the Bagging optimization and integration algorithm vertically to integrate the weak classifiers of each category into a strong classifier.

[0023] In the aforementioned big data testing method, the weak classifier can determine whether the test cases of the horizontal input data under the current score data are accurate, and the strong classifier can determine whether the test cases of all horizontal input data under the current classification data are accurate.

[0024] Compared with the prior art, the beneficial effects of the present invention are:

[0025] Compared with existing technologies, this invention discloses a big data testing system and method based on matrix logic. It includes a big data acquisition module, a big data processing module, a big data testing module, and a big data storage module, all interconnected. It utilizes Hadoop distributed shared storage, the HDFS distributed file system, the HBase distributed database, and the HIVE target data warehouse. In the big data ETL process, Talend is first used to test the first stage of the big data testing—the data verification stage. Secondly, logical verification is performed on a single node, and then this verification is run on multiple nodes to test the second stage of the big data testing—MapReduce. Finally, the results obtained from MapReduce processing in the HDFS distributed file system are used to support the establishment of a high-concurrency distributed database, HBase. For data sources that are relational databases, Sqoop data import functionality is used to migrate them to HBase. The data model of HBase is enhanced by using keywords or timestamps in the sparse sorting mapping table to form data classification, and the classified data is applied to the matrix framework model. By using the Hadoop distributed shared storage principle, big data is acquired, processed, tested, and stored. The third phase of the big data testing used matrix logic to test the accuracy, uniqueness, completeness, consistency, relevance, and timeliness of the data, ensuring data quality and enabling the data to be better and more accurately applied to business decision analysis and recommendation systems.

[0026] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0028] Figure 1 This is an overall architecture diagram of the present invention;

[0029] Figure 2 This is a schematic diagram of MapReduce parallel computing data processing according to the present invention;

[0030] Figure 3 This is a schematic diagram of the matrix logic structure of the third stage of big data testing in this invention;

[0031] Figure 4 This is a schematic diagram of the three-stage testing process for big data testing in this invention;

[0032] Figure 5 This is a schematic diagram illustrating how accurate data is synchronized to the target data warehouse HIVE after the big data test of this invention is completed. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] like Figure 1-4 As shown, a big data testing system based on matrix logic includes a big data acquisition module for collecting data from data sources, including database acquisition, system log acquisition, network data acquisition, and sensor device data acquisition. Data sources include sensors, smart devices, online systems, offline systems, social networks, and internet platforms. The system also includes a big data processing module for processing the ETL process of the data, a big data testing module for testing the ETL process of the data, and a HIVE big data storage module for synchronizing the modified and repaired data after the big data testing to the target data warehouse. All modules are interconnected and communicate with each other.

[0035] Big Data Processing Module: Distributed shared storage is deployed using Hadoop (an open-source framework) on the Master (master-slave) device. YARN (Yet Another Resource Negotiator, a new Hadoop resource manager) is used within Hadoop for global resource management and task scheduling. The distributed shared storage uses the HDFS (Hadoop Distributed File System) distributed file system, leveraging HDFS as the underlying file system to support the generation of HBase (a distributed, column-oriented open-source database)—a distributed key-value store system capable of high concurrency. During the big data ETL (Extract-Transform-Load) process, MapReduce (a programming model for parallel computation of large-scale datasets) is used for distributed parallel computing. Data cleaning, integration, transformation, and reduction are divided into structured data and semi-structured / unstructured data. This data is stored on the HDFS distributed file system and the HBase distributed database, which supports real-time computing. For data sourced from relational databases, Sqoop (an open-source tool primarily used for data transfer between Hadoop and traditional databases, capable of importing data from a relational database into Hadoop's HDFS, and vice versa) is used to migrate the data to the HBase system. The specific details of the data cleaning, integration, transformation, and reduction processes described above are as follows:

[0036] 1. In data cleaning, missing data is filled by using regression analysis, Bayesian calculation formulas, or decision trees to infer the maximum possible value of a specific attribute for that record. For noisy data, the Bin method is used to smooth a set of sorted data by utilizing the nearest neighbors of the data points to be smoothed. The sorted data is then distributed into several Bins.

[0037] 2. Data integration mainly addresses issues such as schema integration, redundancy, data value conflicts, and their elimination.

[0038] 3. Data transformation includes summarization and aggregation operations, data generalization using more abstract concepts to replace low-level or data-layer data objects, and data normalization using a decimal transformation normalization method to project relevant attribute data proportionally to a specific small range.

[0039] 4. In data reduction, data compression uses programs to compress data size and data block reduction utilizes simpler parametric and non-parametric models to replace the original data.

[0040] Big Data Testing Module: In the data ETL process, the first stage of big data testing is the data phase testing. Talend (data integration software) is used to test and verify the following: whether the data resources in various aspects are correctly loaded into the Hadoop system, whether the source data matches the data pushed into the Hadoop system, and whether the correct data is extracted and loaded into the correct location.

[0041] The second phase of big data testing, namely the MapReduce (programming model) phase testing, involves the following tests across the entire distributed device system: verifying whether MapReduce runs normally on a single node and multiple nodes; verifying whether the output files are correct; verifying whether the key-value pairs (key-value distributed storage systems offer fast query speeds, large data storage capacity, and support for high concurrency, making them ideal for queries based on primary keys, but not for complex conditional queries) are correctly generated during the MapReduce processing; and verifying whether the data aggregation and merging are correct after the Reduce process.

[0042] The third stage of big data testing, namely output data testing, involves storing the data from the first two testing stages in a distributed HBase. In the HBase distributed database, the data is aggregated and categorized based on key values. A matrix logic framework is used, employing the AdaBoost iterative algorithm horizontally and the Bagging integration optimization algorithm vertically, to test and verify the accuracy, uniqueness, completeness, consistency, correlation, and timeliness of the data. A comprehensive test report is then generated.

[0043] Big data storage module: The big data storage module includes the Hadoop Distributed File System (HDFS), the Hadoop Distributed Database (HBase) which uses the distributed file system as its underlying support, and HIVE, which synchronizes the data to the target data warehouse after the big data test is completed and changes are made.

[0044] The workflow steps for the big data acquisition module, big data processing module, and big data storage module are as follows:

[0045] 1. Determine the data source, analyze the data source, and prepare the data collection program.

[0046] 2. Determine the Master host, install and deploy Hadoop on the Master, map the IP and hostname, and disable the firewall, configuration files, and environment variables to complete the distributed device setup.

[0047] 3. Ensure the Hadoop service is running and complete the process of writing data to the HDFS distributed file system.

[0048] 4. Determine the preprocessing procedures for cleaning, integrating, transforming, and reducing data during the MapReduce process. This involves completing the corresponding Mapper, Reducer, and Job programs.

[0049] 5. After completing the above configuration, apply the data acquisition program, the data writing program to HDFS, and the preprocessing operation program described above.

[0050] 6. Once the data is loaded into HDFS, MapReduce begins processing the data from different sources.

[0051] A big data testing method based on matrix logic includes the following steps:

[0052] S1. During the data ETL process, the first stage of big data testing is data phase testing and verification, which includes: testing and verifying whether the data resources in all aspects are correctly loaded into the Hadoop system, testing and verifying whether the source data matches the data pushed into the Hadoop system, and testing and verifying whether the correct data is extracted and loaded into the correct location.

[0053] S2. During the data ETL process, conduct the second phase of big data testing: MapReduce phase testing and verification, including: during the MapReduce process of data in HDFS, test and verify whether MapReduce runs normally on a single node and multiple nodes in the entire distributed device system, test and verify whether the output file is correct, test and verify whether the key-value pairs in the MapReduce process are correctly generated, and test and verify whether the data aggregation and merging is correct after the Reduce process is completed.

[0054] S3. In the data ETL process, the third stage of big data testing is carried out: output data stage verification test. After the above MapReduce stage test verification, the results of the MapReduce process are obtained, and the result data is written to the distributed file system HDFS. Based on HDFS as the underlying support, these data results are used to generate the high-concurrency distributed database HBase.

[0055] S4. In distributed HBase, classification is based on its data model enhancement coefficient sorting map, which includes row key, column key, or timestamp.

[0056] S5. In the matrix framework model, a matrix framework is created based on the number of categories. The horizontal input of the matrix is ​​the test cases for the data to be tested. The test cases for the data to be tested include the accuracy, uniqueness, completeness, consistency, relevance, and timeliness of the data. The horizontal matrix uses the AdaBoost iterative algorithm to create weak classifiers for the test cases of the current data to be tested. The vertical matrix uses the Bagging optimization and integration algorithm to integrate the weak classifiers of each category into strong classifiers. Each weak classifier can determine whether the test cases of the horizontal input data under the current score data are accurate, and each strong classifier can determine whether the test cases of all horizontal input data under the current category data are accurate. By integrating the strong classifiers, the accuracy of all test cases of all data categories can be determined.

[0057] The Adaboost algorithm described above has the core idea of ​​training different weak classifiers on the same training set. Its adaptability lies in the fact that samples misclassified by the previous basic classifier will be strengthened, and the weighted samples will be used again to train the next basic classifier. At the same time, a new weak classifier is added in each round until a predetermined sufficiently small error rate is reached or a pre-specified maximum number of iterations is reached.

[0058] The eBagging algorithm described above randomly selects n samples from the original data and repeats this process s times, resulting in s training sets. Each training set can be used to train a classifier, ultimately generating s classifiers. The prediction result is determined by voting among these classifiers, and the class with the most votes is selected as the final prediction result.

[0059] To better understand the technical solution of the present invention, the logical relationships in the accompanying drawings are further explained below:

[0060] Figure 1 In Chinese: The connection between the data source and Hadoop is as follows: After determining the data source, data acquisition method, and data acquisition program, the acquisition program is deployed on the Hadoop Master host. At the same time, Hadoop needs to configure the distributed shared storage device through IP mapping, etc. Hadoop supports that when a shared device fails, no data will be sent to that device. When data files are written to HDFS, HDFS will divide the data into fixed-size data blocks. If the data is smaller than the fixed size, it will maintain its original size. When the data reaches the fixed size of a data block, the HDFS master node will notify the nearest node to request to write the data and return the data block ID and data node to the HDFS client for resource allocation.

[0061] The relationship between Hadoop and YARN: YARN is the manager of Hadoop's resource scheduling and task allocation. It separates resource management and task scheduling monitoring into independent processes. Specifically, YARN includes global resource management and task scheduling, resource management and monitoring of individual nodes, resource management and monitoring of individual jobs, resource request units, and running containers. The first three are the allocation and management of computing resources, while the latter completes the execution of applications.

[0062] The connection between Hadoop and Phase 1 testing: Phase 1 testing mainly verifies whether the data has been correctly loaded into Hadoop. After loading the data, Talend can be used to automatically analyze the data, allowing you to view the data configuration column by column, and also to clean and transform the data type using Talend.

[0063] The connection between MapReduce and second-order testing is as follows: Second-order testing primarily verifies whether the Map and Reduce operations are running correctly, whether the implemented data aggregation and isolation rules are correct, and whether the data key-value pairs are generated correctly. The application is validated by performing these three verifications on a single node and then migrating to multiple nodes for further validation. The output results from the single-node and multi-node tests are then compared to complete this stage of testing.

[0064] The connection between Phase 2 testing, HDFS, and distributed HDFS is as follows: After Phase 2 testing, data is written to HDFS and divided into fixed-size data blocks for storage in the distributed file system HDFS. HDFS also backs up the data blocks to three copies by default. Load balancing and data sharing are achieved among the various HDFS instances in the distributed HDFS cluster.

[0065] The connection between HBase and HDFS is as follows: HBase is the distributed database of Hadoop. HBase is built on HDFS as the underlying system support. It is a distributed, high-concurrency, column-oriented database stored in HDFS. HBase is different from traditional relational databases. It adopts a data model-enhanced sparse sorted mapping table, where the key is composed of row key, column key and timestamp. Classifying by key is the main preparatory work before the three-stage test.

[0066] The connection between classification operations and three-stage testing is as follows: Three-stage testing primarily involves testing the output data, specifically testing the results of MapReduce. It is the final step in big data testing and the last line of defense to ensure data quality. Before three-stage testing, the data in the distributed HBase needs to be integrated, statistically classified, and categorized according to the keywords mentioned above. The more detailed the classification, the better, unless otherwise specified. The categorized data is then input into a matrix model. The matrix model horizontally inputs the data test attributes, such as data accuracy and data completeness. The matrix model contains many weak classifiers. Each weak classifier judges the test attributes of one data point in a category. Horizontally, all classifiers judge one data point in all categories; vertically, all data points in a category are judged; and the combined horizontal and vertical analysis judges all data points in all categories. A test report is then generated by combining the results of the individual and combined tests.

[0067] The connection between the three-stage tests and the target data warehouse HIVE is as follows: After batch processing of the data based on the results of the three-stage tests, the data is stored in the target data warehouse HIVE. HIVE provides SQL query functionality, which can convert SQL statements into distributed parallel computing MapReduce tasks. The stored data can also be written from HIVE to a new HBase. The data in HIVE can be used for statistical queries, for analysis of decision-making directions, or to build data models for personalized recommendation systems.

[0068] Figure 2 In Chinese: The relationship between data and MapReduce: Data enters MapReduce and is processed by the MapReduce program, which is divided into three basic processes: writing Mapper, Reducer, and Job. The Mapper program processes the data, such as data cleaning and data reduction. The Reducer program performs data reduction operations. The Job program executes the above operations. In a distributed system, each MapReduce instance completes the above operations through parallel computing.

[0069] Figure 3In the process, after data is classified, it is output into a matrix model. The horizontal axis of the matrix uses the AdaBoost iterative algorithm to create weak classifiers for testing, with the horizontal input being the data test attributes. The vertical axis uses the Bagging algorithm to integrate the vertical tests. The matrix model is trained using a model trained on a small portion of the correctly classified data. The adaptive nature of AdaBoost lies in strengthening samples misclassified by the previous base classifier. The weighted samples are then used to train the next base classifier. Simultaneously, a new weak classifier is added in each round until a predetermined sufficiently low error rate or a pre-specified maximum number of iterations is reached. This iterative method tests the data test attributes, improving the accuracy of the test results. Bagging integrates and optimizes the vertical axis, improving the overall test results. Testing the entire matrix completes the multi-class data testing task. Each test node and the overall test content are output to a test report, making abnormal data more intuitive and improving data accuracy and quality.

[0070] Figure 4 In China, big data testing is divided into three phases. The first phase, also known as the Pre-Hadoop phase or data verification phase, primarily tests whether various data sources load correct data into the system and whether the data matches the source data. Talend is used to perform these tests. The second phase, MapReduce phase testing, primarily tests whether MapReduce tasks function correctly, whether output configuration files are correct, and whether data key-value pairs are correctly generated. This is done by comparing the results of single-node and synchronous multi-node runs. The third phase, output data testing, primarily tests the accuracy, completeness, and timeliness of the data results processed by MapReduce. This test first uses HDFS as the underlying support to build a distributed database HBase, then classifies the data based on the keywords in the sparse sorting mapping table added to the HBase data model, and finally inputs the classified data into a matrix model and generates a test report to complete the test.

[0071] like Figure 5 As shown, the specific steps for big data storage are as follows:

[0072] 1. Store the collected data, processed by the MapReduce big data processing algorithm, in the distributed file system HDFS;

[0073] 2. After the data in HDFS has passed through the first two testing phases of the big data test, it is stored in HBase, a distributed database supported by HDFS.

[0074] 3. Store the data in HBase, after passing the third stage of the big data test described above and modifying it according to the rules, into the target data warehouse HIVE.

[0075] Based on the results of the three-phase test and the report, the abnormal data was processed according to the established rules and procedures. The processed data was written to a new HDFS file system, and a target data warehouse, HIVE, was created based on HDFS. HIVE maps data files to a database table and provides SQL query functionality, converting SQL statements into MapReduce tasks for execution. Simultaneously, HIVE data can be written to HBase. HIVE is used for statistical queries, while HBase is used for real-time queries. Through real-time querying and statistical analysis, accurate big data can be leveraged more effectively, finding full application in real-time data statistics, decision-making analysis, and customized recommendation systems.

[0076] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. The embodiments selected and specifically described in this specification are intended to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A big data testing method based on matrix logic, characterized in that, A big data testing system based on matrix logic is adopted. The big data testing system includes a big data acquisition module, a big data processing module, a big data testing module, and a big data storage module, and these modules are interconnected. The big data acquisition module is used to collect data from data sources; the big data processing module is used to process the ETL process of the data; the big data testing module is used to test the ETL process of the data; and the big data storage module is used to synchronize the data after the big data testing is completed and the modified and repaired data to the target data warehouse in HIVE. Specifically, the following steps are included: S1. During the data ETL process, conduct the first stage of big data testing: data phase testing and verification. S2. During the data ETL process, conduct the second phase of big data testing: MapReduce phase testing and verification; S3. In the data ETL process, the third stage of big data testing is carried out: output data stage verification test. After the above MapReduce stage test verification, the results of the MapReduce process are obtained, and the result data is written to the distributed file system HDFS. Based on HDFS as the underlying support, these data results are used to generate the high-concurrency distributed database HBase. S4. In distributed HBase, classification is based on its data model-enhanced coefficient sorting mapping table, including row key, column key, or timestamp. S5. In the matrix frame model, create a matrix frame based on the number of categories. The horizontal input of the matrix in the matrix frame is the test cases of the data to be tested. In step S5, the matrix horizontally uses the AdaBoost iterative algorithm to create weak classifiers for the test cases of the current test data; the matrix vertically uses the Bagging optimization fusion algorithm to vertically integrate the weak classifiers of each category into a strong classifier.

2. The big data testing method based on matrix logic as described in claim 1, characterized in that: The data sources include sensors, smart devices, online systems, offline systems, social networks, and internet platforms; the data acquisition methods include database acquisition, system log acquisition, network data acquisition, and sensor device data acquisition.

3. The big data testing method based on matrix logic as described in claim 1, characterized in that: The big data processing module includes online processing of data extraction, transformation, and loading, and data cleaning, integration, transformation, and reduction during MapReduce data processing.

4. The big data testing method based on matrix logic as described in claim 1, characterized in that: The big data testing module includes three testing phases: Phase 1: Data phase verification test; Phase 2: MapReduce phase test; Phase 3: Output data phase verification test, and the modified and repaired data is synchronized to the big data storage module after the big data test is completed.

5. The big data testing method based on matrix logic as described in claim 1, characterized in that: In step S1, the first phase of testing and verification includes: testing and verifying whether the data resources in various aspects are correctly loaded into the Hadoop system, testing and verifying whether the source data matches the data pushed into the Hadoop system, and testing and verifying whether the correct data is extracted and loaded into the correct location.

6. The big data testing method based on matrix logic as described in claim 1, characterized in that: In step S2, the second phase of testing and verification includes: during the MapReduce processing of data in HDFS, testing and verifying whether MapReduce runs normally on a single node and multiple nodes throughout the distributed device system; testing and verifying whether the output files are correct; testing and verifying whether the key-value pairs in the MapReduce processing are correctly generated; and testing and verifying whether the data aggregation and merging are correct after the Reduce process is completed.

7. The big data testing method based on matrix logic as described in claim 1, characterized in that: In step S5, the test cases for the data to be tested include the accuracy, uniqueness, completeness, consistency, relevance, and timeliness of the data.

8. The big data testing method based on matrix logic as described in claim 7, characterized in that: The weak classifier can determine whether the test cases of the horizontal input data under the current score data are accurate, and the strong classifier can determine whether the test cases of all horizontal input data under the current classification data are accurate.

Images