An intelligent survey data analysis method and system based on AI computing power

Abstract

The present invention discloses an intelligent survey data analysis method and system based on AI computing power, which relates to the field of survey data analysis technology, in order to solve the problem of poor data analysis effect. The present invention can reveal the information in the knowledge graph from multiple angles through multi-dimensional analysis, providing more comprehensive and in-depth insights. Each model can automatically run according to the set algorithm and parameters, reducing manual intervention, improving analysis efficiency and accuracy, selecting a channel with a remaining capacity greater than the reported transmission volume for transmission, making full use of transmission channel resources, avoiding resource waste, and also improving overall transmission efficiency. By identifying entity relationships based on grammatical structure, semantic information and context, the present invention fully considers the various characteristics of the data and can more accurately identify the relationship between entities.

Description

Technical Field

[0001] The present invention relates to the field of survey data analysis technology, and specifically to an intelligent survey data analysis method and system based on AI computing power. Background Art

[0002] Survey data analysis refers to the process of using various analytical methods and techniques to process, analyze and interpret data collected through surveys in order to extract valuable information, discover patterns, relationships and trends in the data, and thus provide a basis for decision-making.

[0003] Chinese patent publication number CN117670421A discloses a market research data analysis system based on big data analysis. The system mainly comprises a target area setting module, a market data acquisition module, a data preprocessing module, a market trend monitoring module, a consumer behavior monitoring module, a marketing effect monitoring module, a research data modeling module, and a comprehensive analysis and evaluation module. The target area setting module divides the research target sub-areas, the market data acquisition module collects primary market data of each target sub-area, the data preprocessing module processes and integrates the data to obtain preprocessed secondary data, the market trend monitoring module calculates the market feasibility coefficient, the consumer behavior monitoring module calculates the consumption capacity coefficient, the marketing effect monitoring module calculates the marketing positive feedback coefficient, the research data modeling module calculates the comprehensive market evaluation index, and the comprehensive analysis and evaluation module determines the market risk based on the comprehensive market evaluation index. Although the above patent solves the problem of data analysis, the following problems still exist in actual operation:

[0004] 1. Targeted data analysis is not performed based on the type of each survey data, resulting in poor security when constructing the knowledge graph of the data.

[0005] 2. The survey data were not processed using different processing models, resulting in reduced data mining capabilities.

[0006] 3. After the survey data analysis was completed, a more intuitive report was not generated, and the transmission efficiency was not processed when the generated report was transmitted to the staff, resulting in poor data processing capabilities. Summary of the Invention

[0007] The purpose of the present invention is to provide an intelligent survey data analysis method and system based on AI computing power. Through multi-dimensional analysis, it can reveal the information in the knowledge graph from multiple angles and provide more comprehensive and in-depth insights. Each model can automatically run according to the set algorithms and parameters, reducing manual intervention, improving analysis efficiency and accuracy, and selecting a channel with a remaining capacity greater than the reported transmission volume for transmission, thereby making full use of the transmission channel resources and avoiding resource waste. At the same time, it can also improve the overall transmission efficiency. By identifying entity relationships based on grammatical structure, semantic information and contextual context, it fully considers the various aspects of the data, can more accurately identify the relationship between entities, and can solve the problems in the existing technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] An intelligent survey data analysis method based on AI computing power, including:

[0010] The survey data is collected from different data ports, the collected raw data is processed and fused, the processed and fused raw data is used to construct a knowledge graph, the constructed knowledge graph is intelligently analyzed, and dynamic early warning processing is performed based on the results of the intelligent analysis.

[0011] Preferably, the processed and fused raw data is used to construct a knowledge graph for:

[0012] Before constructing a knowledge graph for raw data, first identify the objectives and domains of the research data analysis. Based on these objectives and domains, determine the types of entity data to be extracted. Entity data types include products, brands, consumers, sales channels, organizations, people, and events. The objectives and domains of the research data analysis are retrieved from the rule database.

[0013] Use natural language processing technology and machine learning algorithms to scan and analyze the processed and integrated raw data;

[0014] The text data in the fused raw data is processed using a named entity recognition algorithm to extract entities from user comments collected from network data ports and offline survey questionnaire text content; the structured data in the fused raw data is processed to extract corresponding entity information from database fields and data labels collected by sensors;

[0015] Relationship recognition is performed on the scanned and analyzed data. Relationship recognition is to identify entity relationships based on the grammatical structure, semantic information, and contextual context in the scanned and analyzed data;

[0016] Verify the identified entity relationships through manual review and comparison across multiple data sources;

[0017] Analyze the characteristics and attribute information of the verified data, add the attribute values ​​extracted from the data to the corresponding entities based on the analysis results, and verify the added attribute values;

[0018] After verification, data extraction is performed on entities, identified relationships, and added attributes, and a knowledge graph is constructed graphically using a knowledge graph construction tool.

[0019] Optimize the constructed knowledge graph, including removing redundant nodes and edges;

[0020] After the optimization is completed, the constructed knowledge graph is obtained.

[0021] Preferably, the collected raw data is processed and fused for:

[0022] The acquired raw data is cleaned, including missing value identification and processing, duplicate data deletion, and erroneous data correction;

[0023] After data cleaning is completed, data conversion is performed, which includes data format unification, data standardization and data coding;

[0024] After the data conversion is completed, data fusion is performed. Data fusion involves analyzing the associated fields between the data obtained from different data ports, and obtaining the association relationship of each associated field after the analysis is completed.

[0025] Integrate data from different data ports into one data set based on the determined association relationships;

[0026] Verify the integrated data set, including data statistical analysis and sampling inspection method;

[0027] After data verification is completed, the processed and fused original data is obtained.

[0028] Preferably, the survey data is collected from different data ports for:

[0029] Different data ports include network data ports, database data ports, sensor data ports, mobile application data ports, and offline data ports;

[0030] Before collecting raw data from each data port, first apply for and confirm the access rights of each port;

[0031] After the application is approved, raw data will be collected from different data ports according to the data collection plan;

[0032] Among them, the data collection plan is retrieved from the rule library;

[0033] After the original data collection is completed, it is stored in the original database.

[0034] Preferably, a data quality assessment is performed on each data port based on its historical operating parameters, and an upper limit value of the data access amount per unit time corresponding to each data port is determined based on the data quality assessment result, for:

[0035] Retrieve the data update frequency deviation from the historical operating parameters of each data port;

[0036] Comparing the data update frequency deviation with a preset deviation reference value;

[0037] When the data update frequency deviation exceeds a preset deviation reference value, the data port whose data update frequency deviation exceeds the preset deviation reference value is used as the target data port;

[0038] Retrieve the time point of each data update corresponding to the target data port;

[0039] Extracting the data information collected within a preset time period after each data update time point from the historical operating parameters of each data port as the target data information; wherein the preset time period corresponds to a time length in the range of 30s-120s;

[0040] Obtaining a data verification failure rate corresponding to each target data port according to target data information corresponding to each target data port;

[0041] Obtaining a data update frequency deviation rate corresponding to the target data port by using a ratio of the data update frequency deviation corresponding to the target data port to a preset deviation reference value;

[0042] The data update frequency deviation rate corresponding to the target data port is obtained by:

[0043] Data update frequency deviation rate = (data update frequency deviation - deviation reference value) / deviation reference value

[0044] Performing difference processing on the data update frequency deviation rate and the data verification failure rate;

[0045] An upper limit value of the data access amount per unit time corresponding to each data port is determined according to the difference between the data update frequency deviation rate and the data verification failure rate.

[0046] Preferably, determining the upper limit of the data access amount per unit time corresponding to each data port according to the difference between the data update frequency deviation rate and the data verification failure rate includes:

[0047] The maximum permissible data access amount per unit time of the target data port is accessed, wherein the value range of the unit time is 1s-5s;

[0048] Retrieve the actual retrieved data volume corresponding to each unit time of the target data port from the historical operation parameters;

[0049] Obtain a reference value of the data access amount per unit time of the target data port according to the actual access data amount corresponding to each unit time and the maximum allowed data access amount;

[0050] Performing a ratio processing on the data amount reference value and the maximum allowable data retrieval amount to obtain a data retrieval reference ratio;

[0051] comparing the difference between the data update frequency deviation rate and the data verification failure rate with a data retrieval reference ratio;

[0052] When the difference between the data update frequency deviation rate and the data verification failure rate is not less than the data retrieval reference ratio, the data retrieval amount per unit time reference value of the target data port is used as the upper limit value of the data retrieval amount per unit time of the target data port;

[0053] When the difference between the data update frequency deviation rate and the data verification failure rate is less than the data retrieval reference ratio, the deviation failure coupling coefficient corresponding to the target data port is retrieved, where the value range of the deviation failure coupling coefficient is [0.4, 0.8];

[0054] The upper limit of the data retrieval amount per unit time of the target data port is set by using the deviation failure coupling coefficient combined with the data update frequency deviation rate and the data verification failure rate.

[0055] Preferably, the constructed knowledge graph is subjected to intelligent analysis for:

[0056] Retrieve sentiment analysis models, cluster analysis models, and regression prediction models from the model library;

[0057] The sentiment analysis model is to use the sentiment analysis model to perform semantic understanding and sentiment judgment on the text data in the constructed knowledge graph;

[0058] The clustering analysis model is as follows: the entity attribute data in the constructed knowledge graph is imported into the clustering analysis model. The clustering analysis model automatically aggregates similar entities into different categories based on the set algorithm and parameters;

[0059] The regression prediction model is as follows: historical data and related attribute features in the constructed knowledge graph are input into the regression prediction model. The regression prediction model predicts future trends by learning the patterns in the data.

[0060] Integrate the analysis results of each model and establish the correlation between the analysis results;

[0061] Finally, intelligent analysis data is obtained.

[0062] Preferably, dynamic early warning processing is performed based on the intelligent analysis results to:

[0063] According to the objectives and fields of survey data analysis, early warning indicators are screened from intelligent analysis data;

[0064] Among them, early warning indicators include product sales growth rate, proportion of dissatisfied consumers and changes in competitor market share;

[0065] And determine the specific location of each warning indicator data in the intelligent analysis results;

[0066] Retrieve historical warning data from the database, determine the fluctuation range and change pattern of each warning indicator based on the historical warning data, and set warning thresholds based on the fluctuation range and change pattern;

[0067] Among them, the warning thresholds include high risk, medium risk and low risk;

[0068] Compare and analyze intelligent analysis data with the set warning threshold through automated procedures;

[0069] Determine whether the intelligent analysis data reaches or exceeds the set warning threshold based on the comparative analysis results;

[0070] Finally, an early warning response is made based on the judgment results, and the early warning response is the result of dynamic early warning processing.

[0071] Preferably, dynamically adjusting the set warning threshold according to the fluctuation range includes:

[0072] Determine the fluctuation range of each warning indicator based on historical warning data;

[0073] Extracting a maximum value and a minimum value of a fluctuation range from the fluctuation range of the early warning indicator;

[0074] Performing ratio processing on the minimum value of the fluctuation amplitude and the maximum value of the fluctuation amplitude to obtain a fluctuation amplitude ratio parameter;

[0075] Extract the data port obtained from the early warning indicators;

[0076] When the target data port does not exist in the data port obtained by the early warning indicator, the set early warning threshold is not adjusted;

[0077] When a target data port exists in the data ports obtained by the early warning indicator, the data update frequency deviation rate and the deviation failure coupling coefficient corresponding to the target data port are retrieved;

[0078] The data update frequency deviation rate corresponding to the target data port in the data port obtained by the early warning indicator is combined with the deviation failure coupling coefficient to adjust the set early warning threshold.

[0079] An intelligent survey data analysis system based on AI computing power, including:

[0080] Data visualization unit, used for:

[0081] Visualize intelligent analysis results and early warning response data;

[0082] Visualization processing involves confirming the display data and warning corresponding data in the intelligent analysis results. The display data in the intelligent analysis results includes trend data, proportion data, relationship data, and comparison data; the warning response data includes warning indicator status, warning level distribution, and response processing progress.

[0083] Use visualization tools to visualize the display data, and obtain chart data and text data after the visualization is completed;

[0084] Evaluate the effect of the completed visualization data. The evaluation includes: evaluating the clarity, aesthetics, and accuracy of the visualization data, and optimizing it based on the evaluation results.

[0085] After the optimization is completed, visual data of the survey data analysis is obtained.

[0086] Preferably, it also includes:

[0087] Report generation transmission unit for:

[0088] Generate visual reports using report generation software using the visual data from survey data analysis;

[0089] Confirming the report transmission volume of the generated visualization report, wherein the report transmission volume is confirmed according to the size of the visualization report;

[0090] Confirm the remaining capacity of the transmission channels of each data receiving port, where the number of transmission channels of each data receiving port is not less than 4;

[0091] Selecting a transmission channel whose remaining capacity is greater than the reported transmission amount as the transmission channel for visualizing the report and transmitting it to different data receiving ports;

[0092] Finally, the staff receives and views the visualization report at different data receiving ports.

[0093] Compared with the prior art, the present invention has the following beneficial effects:

[0094] 1. The intelligent survey data analysis method and system based on AI computing power provided by the present invention extract entities through different methods, can comprehensively mine the information in the data, and recognize entity relationships through grammatical structure, semantic information and contextual context. It fully considers the multi-faceted characteristics of the data and can more accurately identify the relationship between entities. It uses knowledge graph construction tools to construct knowledge graphs in a graphical manner, and displays complex data in intuitive graphics, which is convenient for users to understand and analyze data and helps to discover potential associations and patterns in the data.

[0095] 2. The intelligent survey data analysis method and system based on AI computing power provided by the present invention utilizes automated programs to compare and analyze intelligent analysis data with set warning thresholds, thereby reducing manual intervention, improving the timeliness and accuracy of warnings, and enabling rapid discovery of problems and timely responses. Multi-dimensional analysis can reveal information in the knowledge graph from multiple perspectives, providing more comprehensive and in-depth insights. Each model can automatically run according to the set algorithms and parameters, reducing manual intervention and improving analysis efficiency and accuracy.

[0096] 3. The intelligent survey data analysis method and system based on AI computing power provided by the present invention can help decision makers obtain key information more quickly and accurately by visualizing intelligent analysis results and early warning response data, thereby making decisions more efficiently. Channels with remaining capacity greater than the reported transmission volume are selected for transmission, fully utilizing transmission channel resources and avoiding resource waste, while also improving overall transmission efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 A schematic diagram of the steps of intelligent survey data analysis of the present invention;

[0098] Figure 2 Schematic diagram of the process of intelligent survey data analysis of the present invention. DETAILED DESCRIPTION

[0099] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0100] In order to solve the problem in existing technologies that there is no targeted data analysis based on the type of each survey data, which leads to poor security when constructing knowledge graphs based on data, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0101] An intelligent survey data analysis method based on AI computing power, including:

[0102] The survey data is collected from different data ports, the collected raw data is processed and fused, the processed and fused raw data is used to construct a knowledge graph, the constructed knowledge graph is intelligently analyzed, and dynamic early warning processing is performed based on the results of the intelligent analysis.

[0103] Specifically, raw data collection from different data sources can gather rich and diverse data covering multiple aspects and dimensions, making the survey data more comprehensive, reducing data gaps and biases, and providing a more solid data foundation for subsequent analysis. Processing and fusing the collected raw data can eliminate noise, errors, and inconsistencies, improve data accuracy and consistency, and enhance data quality, providing reliable data support for knowledge graph construction and subsequent analysis. By building a knowledge graph, scattered data can be structured and correlated to form an organic knowledge system. Knowledge graphs can clearly display the relationships and inherent logic between data, helping to better understand survey data and discover hidden patterns and regularities. Leveraging AI computing power to intelligently analyze knowledge graphs, they can automatically mine complex information and potential connections within the data, uncovering trends, patterns, and anomalies that are difficult for humans to detect. AI algorithms can rapidly process large amounts of data, improving analytical efficiency and accuracy, and providing a more forward-looking and scientific basis for decision-making. Dynamic early warning processing based on intelligent analysis results can promptly identify potential problems and risks, enabling relevant personnel to take appropriate preventive and response measures before problems arise, reducing losses and risks, and improving the timeliness and effectiveness of decision-making.

[0104] The processed and integrated raw data is used to construct a knowledge graph for:

[0105] Before constructing a knowledge graph for raw data, first identify the objectives and domains of the research data analysis. Based on these objectives and domains, determine the types of entity data to be extracted. Entity data types include products, brands, consumers, sales channels, organizations, people, and events. The objectives and domains of the research data analysis are retrieved from the rule database.

[0106] Use natural language processing technology and machine learning algorithms to scan and analyze the processed and integrated raw data;

[0107] The text data in the fused raw data is processed using a named entity recognition algorithm to extract entities from user comments collected from network data ports and offline survey questionnaire text content; the structured data in the fused raw data is processed to extract corresponding entity information from database fields and data labels collected by sensors;

[0108] Relationship recognition is performed on the scanned and analyzed data. Relationship recognition is to identify entity relationships based on the grammatical structure, semantic information, and contextual context in the scanned and analyzed data;

[0109] Verify the identified entity relationships through manual review and comparison across multiple data sources;

[0110] Analyze the characteristics and attribute information of the verified data, add the attribute values ​​extracted from the data to the corresponding entities based on the analysis results, and verify the added attribute values;

[0111] After verification, data extraction is performed on entities, identified relationships, and added attributes, and a knowledge graph is constructed graphically using a knowledge graph construction tool.

[0112] Optimize the constructed knowledge graph, including removing redundant nodes and edges;

[0113] After the optimization is completed, the constructed knowledge graph is obtained.

[0114] Specifically, before building the knowledge graph, we first confirm the goals and fields of the survey data analysis, and determine the entity data type based on the goals and fields, so that the construction of the knowledge graph is more targeted and can be carried out closely around specific needs, avoiding waste of resources and inapplicable results caused by blind construction. It processes both text data and structured data, covering data from multiple sources such as user comments collected by network data ports, questionnaire text content of offline surveys, database fields and data labels collected by sensors. By extracting entities through different methods, we can fully mine the information in the data, and identify entity relationships based on grammatical structure, semantic information and context. It fully considers the various characteristics of the data, can more accurately identify the relationship between entities, and improve the accuracy and reliability of the relationships in the knowledge graph. It uses manual review and By comparing multiple data sources, the identified entity relationships are verified. The double verification mechanism helps to ensure the authenticity and consistency of the data, reduce errors and deviations, analyze the characteristics and attribute information of the verified data, and add the attribute values ​​to the corresponding entities. The added attribute values ​​are checked to ensure the accuracy and completeness of the entity attributes in the knowledge graph, so that the knowledge graph can more accurately describe the entities and their relationships. The knowledge graph construction tool is used to construct the knowledge graph in a graphical way, and complex data is displayed in an intuitive graphic to facilitate user understanding and analysis of the data. It helps to discover potential associations and patterns in the data, optimize the constructed knowledge graph, remove redundant nodes and edges, simplify the structure of the knowledge graph, improve the clarity and efficiency of the knowledge graph, and make it easier to use and maintain.

[0115] The collected raw data is processed and integrated for:

[0116] The acquired raw data is cleaned, including missing value identification and processing, duplicate data deletion, and erroneous data correction;

[0117] After data cleaning is completed, data conversion is performed, which includes data format unification, data standardization and data coding;

[0118] After the data conversion is completed, data fusion is performed. Data fusion involves analyzing the associated fields between the data obtained from different data ports, and obtaining the association relationship of each associated field after the analysis is completed.

[0119] Integrate data from different data ports into one data set based on the determined association relationships;

[0120] Verify the integrated data set, including data statistical analysis and sampling inspection method;

[0121] After data verification is completed, the processed and fused original data is obtained.

[0122] It has the advantages of removing noise, errors and inconsistencies in the original data through data cleaning operations such as missing value identification and processing, duplicate data deletion and erroneous data correction, improving the accuracy and completeness of the data, and providing a reliable data foundation for subsequent analysis and application. The data format is unified, standardized and coded in the data conversion process, making the data consistent and standardized, facilitating data sharing and interaction between different systems or departments, and also helping to improve the efficiency of data processing and analysis. In the data fusion stage, the associated fields between the data obtained from different data ports are analyzed to discover the potential relationships and patterns between the data, which helps to have a more comprehensive understanding of the business or phenomenon represented by the data, and provide richer information support for decision-making. Integrating them into one data set enables centralized management and integrated utilization of data, avoids the problem of information islands caused by data scattered across different systems or departments, facilitates comprehensive analysis and mining, and thus better unleashes the value of data. Through data statistical analysis and sampling inspection methods for data verification, a comprehensive quality assessment of the processed and integrated data can be conducted, and possible problems can be discovered in a timely manner to ensure the availability and reliability of the data, and to ensure the accuracy and effectiveness of decisions and applications based on the data. After data processing and integration, the data set is more standardized, complete and relevant, and can better adapt to different analysis needs and application scenarios. Whether it is used for daily business report generation or complex data analysis and mining model construction, it can provide high-quality data support.

[0123] The survey data is collected from different data ports and used for:

[0124] Different data ports include network data ports, database data ports, sensor data ports, mobile application data ports, and offline data ports;

[0125] Before collecting raw data from each data port, first apply for and confirm the access rights of each port;

[0126] After the application is approved, raw data will be collected from different data ports according to the data collection plan;

[0127] Among them, the data collection plan is retrieved from the rule library;

[0128] After the original data collection is completed, it is stored in the original database.

[0129] Specifically, applying for and confirming access rights for each port before collecting data can ensure that data collection is carried out under the premise of legal authorization, effectively preventing unauthorized access and data leakage, protecting the rights and interests of data owners, and maintaining data security and privacy. Retrieving data collection plans from the rule library provides clear and standardized operating procedures and standards for data collection, ensuring the consistency and accuracy of the collection process, helping to improve data quality and reduce data deviations and errors caused by human factors or inconsistent operations. Covering multiple data ports such as networks, databases, sensors, mobile applications, and offline, it can comprehensively collect data from different sources and integrate multi-channel data, thereby more comprehensively and richly reflecting the characteristics and behaviors of the survey subjects and providing more comprehensive data support for subsequent analysis and decision-making. After the original data is collected, it is stored in the original database, facilitating centralized management and storage of data, facilitating subsequent data query, retrieval, backup, and maintenance, and helping to improve data management efficiency and data availability. Adopting corresponding collection methods for different data ports can adapt to the characteristics and collection requirements of various types of data. Whether it is structured database data, unstructured network data, or sensor data, it can be effectively collected and processed.

[0130] Specifically, a data quality assessment is performed on each data port based on its historical operating parameters, and an upper limit on the amount of data retrieved per unit time corresponding to each data port is determined based on the data quality assessment result, for:

[0131] Retrieve the data update frequency deviation from the historical operating parameters of each data port;

[0132] Comparing the data update frequency deviation with a preset deviation reference value;

[0133] When the data update frequency deviation exceeds a preset deviation reference value, the data port whose data update frequency deviation exceeds the preset deviation reference value is used as the target data port;

[0134] Retrieve the time point of each data update corresponding to the target data port;

[0135] Extracting the data information collected within a preset time period after each data update time point from the historical operating parameters of each data port as the target data information; wherein the preset time period corresponds to a time length in the range of 30s-120s;

[0136] Obtaining a data verification failure rate corresponding to each target data port according to target data information corresponding to each target data port;

[0137] Obtaining a data update frequency deviation rate corresponding to the target data port by using a ratio of the data update frequency deviation corresponding to the target data port to a preset deviation reference value;

[0138] The data update frequency deviation rate corresponding to the target data port is obtained by:

[0139] Data update frequency deviation rate = (data update frequency deviation - deviation reference value) / deviation reference value

[0140] Performing difference processing on the data update frequency deviation rate and the data verification failure rate;

[0141] An upper limit value of the data access amount per unit time corresponding to each data port is determined according to the difference between the data update frequency deviation rate and the data verification failure rate.

[0142] The technical effect of the above technical solution is: by retrieving the data update frequency deviation and comparing it with the preset deviation reference value, the data port with abnormal data update frequency can be accurately found. At the same time, the data verification failure rate is calculated by combining the data information of each data update time point of the target data port and the data information within the preset time period, and the data quality is comprehensively evaluated from the two dimensions of data update frequency and data verification, so that the evaluation results are more comprehensive and accurate. The upper limit of the data retrieval volume per unit time corresponding to each data port is determined based on the difference between the data update frequency deviation rate and the data verification failure rate, and the data retrieval volume can be dynamically adjusted according to the data quality. For data ports with poor data quality, the upper limit of the data retrieval volume per unit time is reduced to avoid the system processing too much invalid or erroneous data due to data quality problems, thereby improving the overall data processing efficiency and performance stability of the system. Prevent the operation of the entire system from being affected by the poor data quality of individual data ports, ensure the stability of the system in data collection, processing and transmission, reduce system failures and errors caused by data problems, and improve the reliability and availability of the system. Based on the data quality assessment results, an upper limit on the data retrieval volume is allocated so that system resources (such as network bandwidth, storage resources, computing resources, etc.) can be allocated first to data ports with high data quality, thereby achieving rational use of resources, avoiding resource waste, and improving the overall utilization of system resources.

[0143] Specifically, determining the upper limit of the data access amount per unit time corresponding to each data port according to the difference between the data update frequency deviation rate and the data verification failure rate includes:

[0144] The maximum permissible data access amount per unit time of the target data port is accessed, wherein the value range of the unit time is 1s-5s;

[0145] Retrieve the actual retrieved data volume corresponding to each unit time of the target data port from the historical operation parameters;

[0146] Obtain a reference value of the data access amount per unit time of the target data port according to the actual access data amount corresponding to each unit time and the maximum allowed data access amount;

[0147] The reference value of the data access amount per unit time of the target data port is obtained by the following formula:

[0148] ;

[0149] Among them, U c Indicates the reference value of the data access amount per unit time of the target data port; n indicates the number of unit time experienced by the target data port; U i Indicates the actual amount of data retrieved corresponding to the i-th unit time of the target data port; U max Indicates the maximum amount of data allowed to be retrieved from the target data port; U p and U z They represent the average value and median value of the actual retrieved data volume corresponding to the unit time of the target data port respectively; The minimum value of the two is taken in order to select a relatively conservative value that is more representative of the general level of the data volume from the overall characteristics of the data volume, to avoid overestimating the data volume due to the average value being affected by extreme large values, and to lay a relatively robust foundation for subsequent calculations. This formula comprehensively considers the situation of the target data port in multiple unit times to measure the status of its actual data retrieval relative to the maximum allowable amount. Compare the actual amount of data retrieved per unit time with the maximum allowable amount of data to obtain the share of the actual amount in the maximum allowable amount. On this basis, the remaining ratio of the actual amount per unit time to the maximum allowable amount is calculated to reflect the degree to which the maximum allowable amount has not been reached. The remaining proportions are nonlinearly transformed and normalized using the hyperbolic tangent function and mapped to the (−1,1) interval. This not only highlights the difference between the actual and maximum allowable amounts of each unit time, but also avoids excessive interference from extreme values. The transformed values ​​of n time units are summed to fully cover the conditions of all time units. Finally, Taking the average of the sums yields an average measure of the degree to which the actual data volume deviates from the maximum allowable volume over multiple time units, accurately measuring the relative state of data access across the data port. By combining the average, median, and actual data volume per time unit, this formula comprehensively considers the overall level, intermediate trends, and specific conditions of the target data port during operation. This allows the calculated data access reference value to more comprehensively and accurately reflect the data access characteristics of the data port, avoiding the bias inherent in relying solely on a single data characteristic (such as the average). Using the tanh function to perform a nonlinear transformation on the difference between the actual data volume and the maximum allowable volume per time unit more sensitively captures the degree to which data volume deviates from the maximum allowable volume at different time units, highlighting the impact of these differences when calculating the reference value. This allows the reference value to better reflect the actual fluctuations in data access at the data port and improves its fit to actual data access conditions. Taking the minimum of the average and median values ​​ensures robustness and reduces the risk of interference from extreme data. At the same time, through comprehensive calculation and processing of the data volume per unit time, it can adapt to the data retrieval modes and fluctuations of different data ports, so that the calculated data retrieval volume reference value has good representativeness and applicability in different scenarios, providing a reliable basis for the subsequent determination of the upper limit of the data retrieval volume.

[0150] Performing a ratio processing on the data amount reference value and the maximum allowable data retrieval amount to obtain a data retrieval reference ratio;

[0151] comparing the difference between the data update frequency deviation rate and the data verification failure rate with a data retrieval reference ratio;

[0152] When the difference between the data update frequency deviation rate and the data verification failure rate is not less than the data retrieval reference ratio, the data retrieval amount per unit time reference value of the target data port is used as the upper limit value of the data retrieval amount per unit time of the target data port;

[0153] When the difference between the data update frequency deviation rate and the data verification failure rate is less than the data retrieval reference ratio, the deviation failure coupling coefficient corresponding to the target data port is retrieved, where the value range of the deviation failure coupling coefficient is [0.4, 0.8];

[0154] The upper limit of the data retrieval amount per unit time of the target data port is set by using the deviation failure coupling coefficient combined with the data update frequency deviation rate and the data verification failure rate;

[0155] The upper limit of the data access amount is obtained by the following formula:

[0156] ;

[0157] Among them, U up represents the upper limit of data retrieval; F represents the data verification failure rate; D represents the data update frequency deviation rate; x represents the deviation failure coupling coefficient; y represents the failure rate correction factor, and its value range is 0.7≤β≤1.2. middle By correlating the data verification failure rate with the data update frequency deviation rate, we can measure the combined impact of these two factors on data quality. We then transform this using the logarithmic function ln. The characteristic of logarithmic functions is that they change slowly when the independent variable is small and dramatically when the independent variable is large. This allows us to nonlinearly quantify the combined impact of the data verification failure rate and the data update frequency deviation rate, making their impact on the final result more consistent with the actual effects of data quality factors. The exponential function exp is used to calculate the product of the data update frequency deviation rate and the data verification failure rate. Due to the exponential function's decaying nature, its value decreases rapidly as D and F increase. This reflects the negative impact of both data update frequency deviation and data verification failure on data quality, which increases rapidly as their values ​​increase. Taking the minimum value of the logarithmic and exponential terms is to select the result that more accurately reflects the combined impact of data quality, from the two different quantification methods. Based on the quantitative value of the combined effect of data verification failure rate and data update frequency deviation rate calculated above, 1 is added to ensure that the result of this part of the calculation is greater than 1, so that c When multiplied, it can be c Based on the data quality, the upper limit of data retrieval should be reasonably adjusted so that it will not be less than U c This is in line with the logic that the upper limit of data retrieval volume in actual needs should be based on the reference value and reasonably changed according to the data quality (generally adjusted based on the reference value).

[0158] The technical effect of the above technical solution is as follows: by considering factors such as the maximum allowable data retrieval amount per unit time of the target data port and the actual data retrieval amount, a data retrieval amount reference value and a data retrieval reference ratio are calculated, and compared with the difference between the data update frequency deviation rate and the data verification failure rate. This allows for precise determination of the upper limit of the data retrieval amount per unit time for each data port based on the actual conditions of each data port, thereby achieving refined management of data retrieval. When the difference between the data update frequency deviation rate and the data verification failure rate is not less than the data retrieval reference ratio, the data retrieval amount reference value is used as the upper limit, thereby preventing data ports with poor data quality from excessively occupying resources. When the difference is less than the reference ratio, the upper limit is set by introducing a deviation failure coupling coefficient, combining the data update frequency deviation rate and the data verification failure rate. This allows for more flexible adjustment of resource allocation based on data quality and actual retrieval conditions, thereby improving the utilization efficiency of system resources and ensuring that resources are preferentially allocated to data ports with better data quality. The setting of the deviation-failure coupling coefficient enables the system to dynamically adjust the upper limit of data retrieval based on different data quality conditions and data retrieval scenarios, enhancing the system's adaptability to various complex data scenarios. This dynamic adjustment mechanism helps maintain system stability, preventing system overloads and crashes caused by data quality issues or data retrieval anomalies at data ports, thereby improving system reliability and stability. Determining the upper limit of data retrieval based on the relationship between the data update frequency deviation rate and the data verification failure rate, as well as actual data retrieval conditions, ensures that the data processed by the system is more accurate and reliable. By reasonably limiting the data retrieval volume from data ports with data quality issues, the impact of erroneous or incomplete data on system processing results is reduced, thereby improving the accuracy and effectiveness of data processing across the entire system.

[0159] In order to solve the problem that the existing technology does not use different processing models to process survey data, which leads to reduced data mining capabilities, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0160] The constructed knowledge graph is intelligently analyzed for:

[0161] Retrieve sentiment analysis models, cluster analysis models, and regression prediction models from the model library;

[0162] The sentiment analysis model is to use the sentiment analysis model to perform semantic understanding and sentiment judgment on the text data in the constructed knowledge graph;

[0163] The clustering analysis model is as follows: the entity attribute data in the constructed knowledge graph is imported into the clustering analysis model. The clustering analysis model automatically aggregates similar entities into different categories based on the set algorithm and parameters;

[0164] The regression prediction model is as follows: historical data and related attribute features in the constructed knowledge graph are input into the regression prediction model. The regression prediction model predicts future trends by learning the patterns in the data.

[0165] Integrate the analysis results of each model and establish the correlation between the analysis results;

[0166] Finally, intelligent analysis data is obtained.

[0167] Specifically, it can fully exploit the value of different types of data within the knowledge graph. Whether it's text data, entity attribute data, or historical data, each can be deeply analyzed using corresponding models, avoiding data waste and maximizing the utilization of the knowledge graph's information. Sentiment analysis models can understand the sentiment within text, providing support for user feedback and other aspects. Cluster analysis models can identify similarities and inherent structures between entities, facilitating classification management and knowledge discovery. Regression prediction models can predict future trends based on historical data to aid decision-making. This multi-dimensional analysis reveals information within the knowledge graph from multiple perspectives, providing more comprehensive and in-depth insights. Each model automatically runs according to predefined algorithms and parameters, reducing manual intervention and improving analytical efficiency and accuracy. Furthermore, the model can automatically learn patterns within the data, adapting to different data characteristics and variations, demonstrating strong intelligent capabilities. By integrating the analysis results of each model and establishing relationships between them, a more complete and systematic knowledge system can be formed. This correlation analysis can uncover potential connections between different analysis results, uncovering more valuable information and providing stronger support for solving complex problems and making decisions. The resulting intelligent analysis data can provide a scientific basis for various decisions. For example, sentiment analysis can help companies understand the public's attitude towards products or services, cluster analysis helps with market segmentation and target customer positioning, and regression prediction can provide forward-looking information for resource planning, market strategy adjustments, etc., thereby helping companies or organizations better respond to market changes and improve their competitiveness. Among them, integration methods include data layer integration, feature layer integration, and result layer integration. Data layer integration uses the entity's unique identifier as the key to associate sentiment analysis results, cluster labels, and predicted data to the same entity dimension table. Feature layer integration converts multi-model outputs into feature vectors and inputs them into the downstream comprehensive analysis model. Result layer integration uses visualization tools to display multi-model conclusions in parallel and manually analyze cross-correlations. Correlation analysis methods include causal analysis, co-occurrence analysis, and spatiotemporal correlation analysis. Causal analysis uses regression model results to verify the correlation between sentiment tendencies and sales changes. Co-occurrence analysis identifies the co-occurrence patterns of cluster group characteristics and sentiment keywords through association rule mining. Spatiotemporal correlation analysis combines time series data with spatial data to analyze regional differences in trend changes.

[0168] Perform dynamic early warning processing based on intelligent analysis results, used for:

[0169] According to the objectives and fields of survey data analysis, early warning indicators are screened from intelligent analysis data;

[0170] Among them, early warning indicators include product sales growth rate, proportion of dissatisfied consumers and changes in competitor market share;

[0171] And determine the specific location of each warning indicator data in the intelligent analysis results;

[0172] Retrieve historical warning data from the database, determine the fluctuation range and change pattern of each warning indicator based on the historical warning data, and set warning thresholds based on the fluctuation range and change pattern;

[0173] Among them, the warning thresholds include high risk, medium risk and low risk;

[0174] Compare and analyze intelligent analysis data with the set warning threshold through automated procedures;

[0175] Determine whether the intelligent analysis data reaches or exceeds the set warning threshold based on the comparative analysis results;

[0176] Finally, an early warning response is made based on the judgment results, and the early warning response is the result of dynamic early warning processing.

[0177] Specifically, based on the objectives and areas of survey data analysis, it can accurately screen out key early warning indicators related to product sales, consumer feedback, and competitors, making early warnings more targeted and directly reflecting the key influencing factors of the business. By retrieving historical early warning data from the database and analyzing the fluctuation range and change patterns of each early warning indicator to set the early warning threshold, data-based decision-making makes the setting of early warning thresholds more scientific and reasonable, avoiding the limitations of subjective judgment. The early warning thresholds are divided into three levels: high risk, medium risk, and low risk, which can manage and respond to different degrees of risks in a more detailed manner, helping enterprises to take corresponding response measures according to different risk levels and improve the efficiency and effectiveness of risk management. The use of automated programs to compare and analyze intelligent analysis data with set early warning thresholds reduces manual intervention, improves the timeliness and accuracy of early warnings, can quickly identify problems and respond in a timely manner, and help enterprises adjust their strategies in a timely manner to cope with market changes. Dynamic early warning processing is carried out based on the results of intelligent analysis. As the market environment and business data continue to change, continuous monitoring and analysis of data and timely adjustment of early warning responses enable enterprises to better adapt to the dynamically changing market environment and maintain their competitive advantage.

[0178] Specifically, the set warning threshold is dynamically adjusted according to the fluctuation range, including:

[0179] Determine the fluctuation range of each warning indicator based on historical warning data;

[0180] Extracting a maximum value and a minimum value of a fluctuation range from the fluctuation range of the early warning indicator;

[0181] Performing ratio processing on the minimum value of the fluctuation amplitude and the maximum value of the fluctuation amplitude to obtain a fluctuation amplitude ratio parameter;

[0182] Extract the data port obtained from the early warning indicators;

[0183] When the target data port does not exist in the data port obtained by the early warning indicator, the set early warning threshold is not adjusted;

[0184] When a target data port exists in the data ports obtained by the early warning indicator, the data update frequency deviation rate and the deviation failure coupling coefficient corresponding to the target data port are retrieved;

[0185] The data update frequency deviation rate corresponding to the target data port in the data port obtained by the early warning indicator is combined with the deviation failure coupling coefficient to adjust the set early warning threshold;

[0186] The adjusted warning threshold is obtained through the following formula:

[0187] ;

[0188] Among them, R 01 Represents the warning threshold corresponding to the judgment between high risk and medium risk after adjustment; R x Indicates the warning threshold for judging between high risk and medium risk before adjustment; m indicates the number of target data ports in the data ports obtained by the warning indicator; x i represents the deviation failure coupling coefficient corresponding to the i-th target data port; D i represents the data update frequency deviation rate corresponding to the i-th target data port; D c Indicates the preset data update frequency deviation rate reference value; specifically, middle This represents the weighted deviation of the data update frequency of a single target data port, combined with the coupling weight. Adding 1 to this value ensures that the calculated result is greater than 0, allowing the warning threshold to be adjusted in an appropriate proportional manner, reflecting the impact of a single target data port on the warning threshold adjustment. middle This multiplication operation multiplies the aforementioned impact factors for m target data ports. m is the number of target data ports in the data ports obtained from the warning indicator. This multiplication comprehensively considers the combined effect of all target data ports on the warning threshold adjustment, cumulatively accumulating the impact of each target data port and reflecting the combined effect of data quality factors across multiple target data ports. Raising the multiplication result to the mth power is a normalization method. Because multiplication causes the value to increase rapidly with the number of target data ports m, raising the result to the mth power adjusts the combined impact result to a relatively reasonable range, allowing it to be multiplied by the pre-adjusted warning threshold Rx at an appropriate ratio. This avoids excessive adjustments due to an increase in the number of target data ports, ensuring the rationality and stability of the adjustment. This method considers the data update frequency deviation rate of each target data port and its degree of deviation from the reference value, and combines the deviation failure coupling coefficient to comprehensively adjust the warning threshold. This more accurately reflects the impact of actual data quality changes on risk assessment, ensuring that the warning thresholds for high and medium risk assessments are more aligned with actual conditions, improving the accuracy of risk warnings, and reducing the probability of misjudgment. The warning threshold can be dynamically adjusted based on the number of target data ports and the specific data update frequency deviation rate and coupling coefficient of each port. This allows for appropriate responses to target data ports of varying numbers and data quality, enhancing the warning system's adaptability to complex and changing data environments and enabling its effectiveness in a variety of scenarios. By performing m-th-power normalization, the system avoids excessive or insufficient warning threshold adjustments due to changes in the number of target data ports, ensuring the stability of the warning threshold adjustment process. This prevents the warning system from experiencing abnormal fluctuations due to the number of data ports when processing multiple target data ports, maintaining reliable operation.

[0189] ;

[0190] Among them, R 02 Represents the warning threshold corresponding to the judgment between medium risk and low risk after adjustment; R y Indicates the warning threshold corresponding to the judgment between medium risk and low risk before adjustment. The goal is to obtain a factor less than or equal to 1. This factor decreases as the combined impact of the data update frequency deviation rate and the coupling coefficient increases. It represents the adjustment ratio to the original warning threshold after considering the data quality of a single target data port. The more severe the data quality issue, the smaller the adjustment ratio. middle The above adjustment factors of m target data ports are multiplied together. m is the number of target data ports. Through the multiplication operation, the influence of all target data ports on the warning threshold is accumulated, and the effect of the combined effect of the data quality factors of multiple target data ports is comprehensively reflected. The multiplication result is raised to the power of m to normalize it. As the number of target data ports m increases, the multiplication result will become smaller. The purpose of raising it to the power of m is to adjust the comprehensive influence result to an appropriate range so that it can be appropriately proportional to the warning threshold R before adjustment. y Multiplication avoids excessive or insufficient adjustments due to changes in the number of target data ports. By comprehensively considering the data update frequency deviation rate and the deviation failure coupling coefficient of the target data ports, the impact of data quality on risk assessment can be accurately measured, making the warning thresholds for medium and low risk assessments more consistent with actual conditions, reducing misjudgments, and improving the accuracy of risk warnings. The warning threshold can be dynamically adjusted based on the number of target data ports and the specific data quality status of each port. Regardless of the number of target data ports or the differences in data quality, it can respond flexibly and adapt to different data environments, ensuring that the warning system works effectively in various scenarios. The normalization process of raising the value to the mth power prevents the warning threshold adjustment from getting out of control due to changes in the number of target data ports, ensuring a stable adjustment process, and enabling the warning system to operate reliably when processing multiple target data ports without experiencing abnormal fluctuations due to the number of ports.

[0191] The technical solution achieves the following: By determining the fluctuation range of each warning indicator using historical warning data and extracting the maximum fluctuation amplitude calculation ratio parameter, the fluctuation characteristics of the warning indicator can be accurately grasped. When a target data port is involved, the warning threshold is adjusted based on its data update frequency deviation rate and the deviation failure coupling coefficient. This ensures that the warning threshold is closely aligned with the actual fluctuation characteristics of the data port, improving the adaptability of the warning threshold to data changes, thereby enhancing warning accuracy and reducing false positives and missed negatives. The warning threshold can be dynamically adjusted based on whether the data port is the target data port and related parameters, rather than using a fixed threshold. This allows for flexible response to changes in data port data quality and operational status, enabling the warning system to function effectively in various scenarios and enhancing its adaptability to complex and changing data environments. If the target data port is not among the data ports obtained for the warning indicator, the warning threshold is not adjusted, avoiding meaningless computation and resource consumption. Computing resources are concentrated on the presence of the target data port, ensuring that resources are used for critical scenarios, improving system resource utilization efficiency, and enabling more efficient system operation. Reasonable adjustment of the warning threshold can timely and accurately detect data port anomalies, making it easier for operation and maintenance personnel to intervene in advance to deal with potential problems, prevent data port problems from expanding and affecting the overall operation of the system, help maintain system stability and reliability, and reduce the risk of system failures caused by data anomalies.

[0192] In order to solve the problem in the existing technology that after the survey data analysis is completed, a more intuitive report is not generated, and the generated report is not processed for transmission efficiency when it is transmitted to the staff, resulting in poor data processing capabilities, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0193] An intelligent survey data analysis system based on AI computing power, including:

[0194] Data visualization unit, used for:

[0195] Visualize intelligent analysis results and early warning response data;

[0196] Visualization processing involves confirming the display data and warning corresponding data in the intelligent analysis results. The display data in the intelligent analysis results includes trend data, proportion data, relationship data, and comparison data; the warning response data includes warning indicator status, warning level distribution, and response processing progress.

[0197] Use visualization tools to visualize the display data, and obtain chart data and text data after the visualization is completed;

[0198] Evaluate the effect of the completed visualization data. The evaluation includes: evaluating the clarity, aesthetics, and accuracy of the visualization data, and optimizing it based on the evaluation results.

[0199] After the optimization is completed, visual data of the survey data analysis is obtained.

[0200] Specifically, it can perform targeted visualization processing on different types of display data in intelligent analysis results and early warning response data, such as trend, proportion, relationship, comparison data, as well as early warning indicator status, warning level distribution, response processing progress, etc., so that complex data can be presented in the form of intuitive charts and text, which is convenient for users to quickly understand and grasp the core information of the data. By visualizing intelligent analysis results and early warning response data, it can help decision makers obtain key information more quickly and accurately, so as to make decisions more efficiently, especially in the face of emergencies, they can take corresponding measures in time according to the visualized early warning information. The visualization production tool converts the display data into chart data and text data, which can not only present the surface information of the data, but also help users discover the hidden relationships and problems between the data. For example, through trend data, you can observe the changing trend of the data, discover abnormal situations in time, evaluate and optimize the clarity, beauty and accuracy of the visualized data, and ensure The data visualization results finally presented to users are of high quality, accurately reflecting the actual situation of the data and having good visual effects, which are convenient for users to read and analyze. Different types of data are classified, processed and visualized, which helps users understand the data from multiple perspectives. For example, proportion data can let users understand the proportion of each part in the whole, relationship data can show the degree of correlation between data, and comparison data makes it easier for users to compare data. Among them, clarity assessment is achieved through resolution detection, information density analysis and color contrast verification, aesthetics assessment is optimized through design principles and user experience standards, and accuracy assessment is achieved through data consistency verification, chart type adaptability check and trend restoration analysis. Optimization is carried out according to the evaluation results of clarity, aesthetics and accuracy. The optimization execution process is as follows: call the automated evaluation tool to obtain scores for each dimension, generate optimization plans according to preset rules, use the visualization production tool API to perform optimization, re-evaluate until all indicators meet the standards, and generate the final visualization data.

[0201] Report generation transmission unit for:

[0202] Generate visual reports using report generation software using the visual data from survey data analysis;

[0203] Confirming the report transmission volume of the generated visualization report, wherein the report transmission volume is confirmed according to the size of the visualization report;

[0204] Confirm the remaining capacity of the transmission channels of each data receiving port, where the number of transmission channels of each data receiving port is not less than 4;

[0205] Selecting a transmission channel whose remaining capacity is greater than the reported transmission amount as the transmission channel for visualizing the report and transmitting it to different data receiving ports;

[0206] Finally, the staff receives and views the visualization report at different data receiving ports.

[0207] Specifically, using report generation software to generate visual reports from the visual data of survey data analysis can present data in an intuitive and easy-to-understand manner, making it easier for staff to quickly understand and grasp the survey results and improve data interpretation efficiency. By confirming the report transmission volume and the remaining capacity of the transmission channel of each data receiving port, the appropriate transmission channel can be accurately selected to ensure that the visual report can be smoothly transmitted, avoiding transmission failure or data loss due to insufficient channel capacity. The channel with a remaining capacity greater than the report transmission volume is selected for transmission, which fully utilizes the transmission channel resources and avoids resource waste. At the same time, it can also improve overall transmission efficiency. The number of transmission channels for each data receiving port is no less than 4, which increases the redundancy and reliability of transmission. Even if a transmission channel fails or is congested, it can be transmitted through other channels to ensure that the visual report can successfully reach the data receiving port. The transmission channel can be flexibly selected according to the size of different visual reports and the remaining capacity of the transmission channel of each data receiving port. It is suitable for the transmission of survey data visualization reports of different scales and complexities.

[0208] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0209] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. An intelligent survey data analysis method based on AI computing power, characterized by: include: Collect survey data from different data ports, process and fuse the collected raw data, construct a knowledge graph based on the processed and fused raw data, perform intelligent analysis on the constructed knowledge graph, and perform dynamic early warning processing based on the intelligent analysis results; The survey data is collected from different data ports to: Different data ports include network data ports, database data ports, sensor data ports, mobile application data ports, and offline data ports; Perform data quality assessment on each data port based on its historical operating parameters, and determine the upper limit of the data retrieval volume per unit time for each data port based on the data quality assessment results; Perform data quality assessment on each data port based on its historical operating parameters, and determine the upper limit of the data access volume per unit time for each data port based on the data quality assessment results. This is also used for: Retrieve the data update frequency deviation from the historical operating parameters of each data port; Comparing the data update frequency deviation with a preset deviation reference value; When the data update frequency deviation exceeds a preset deviation reference value, the data port whose data update frequency deviation exceeds the preset deviation reference value is used as the target data port; Retrieve the time point of each data update corresponding to the target data port; Extracting the data information collected within a preset time period after each data update time point from the historical operating parameters of each data port as the target data information; wherein the preset time period corresponds to a time length in the range of 30s-120s; Obtaining a data verification failure rate corresponding to each target data port according to target data information corresponding to each target data port; Obtaining a data update frequency deviation rate corresponding to the target data port by using a ratio of the data update frequency deviation corresponding to the target data port to a preset deviation reference value; The data update frequency deviation rate corresponding to the target data port is obtained by: Data update frequency deviation rate = (data update frequency deviation - deviation reference value) / deviation reference value Performing difference processing on the data update frequency deviation rate and the data verification failure rate; An upper limit value of the data access amount per unit time corresponding to each data port is determined according to the difference between the data update frequency deviation rate and the data verification failure rate.

2. The intelligent survey data analysis method based on AI computing power according to claim 1 is characterized in that: The processed and integrated raw data is used to construct a knowledge graph for: Before constructing a knowledge graph for raw data, first identify the objectives and domains of the research data analysis. Based on these objectives and domains, determine the types of entity data to be extracted. Entity data types include products, brands, consumers, sales channels, organizations, people, and events. The objectives and domains of the research data analysis are retrieved from the rule database. Use natural language processing technology and machine learning algorithms to scan and analyze the processed and integrated raw data; The text data in the fused raw data is processed using a named entity recognition algorithm to extract entities from user comments collected from network data ports and offline survey questionnaire text content; the structured data in the fused raw data is processed to extract corresponding entity information from database fields and data labels collected by sensors; Relationship recognition is performed on the scanned and analyzed data. Relationship recognition is to identify entity relationships based on the grammatical structure, semantic information, and contextual context in the scanned and analyzed data; Verify the identified entity relationships, including through manual review and comparison of multiple data sources; Analyze the characteristics and attribute information of the verified data, add the attribute values ​​extracted from the data to the corresponding entities based on the analysis results, and verify the added attribute values; After verification, data extraction is performed on entities, identified relationships, and added attributes, and a knowledge graph is constructed graphically using a knowledge graph construction tool. Optimize the constructed knowledge graph, including removing redundant nodes and edges; After the optimization is completed, the constructed knowledge graph is obtained.

3. The intelligent survey data analysis method based on AI computing power according to claim 2 is characterized in that: The collected raw data is processed and integrated for: The acquired raw data is cleaned, including missing value identification and processing, duplicate data deletion, and erroneous data correction; After data cleaning is completed, data conversion is performed, which includes data format unification, data standardization and data coding; After the data conversion is completed, data fusion is performed. Data fusion involves analyzing the associated fields between the data obtained from different data ports, and obtaining the association relationship of each associated field after the analysis is completed. Integrate data from different data ports into one data set based on the determined association relationships; Verify the integrated data set, including data statistical analysis and sampling inspection method; After data verification is completed, the processed and fused original data is obtained.

4. The intelligent survey data analysis method based on AI computing power according to claim 3 is characterized in that: The original survey data is collected from different data ports and is also used for: Before collecting raw data from each data port, first apply for and confirm the access rights of each port; After the application is approved, raw data will be collected from different data ports according to the data collection plan; Among them, the data collection plan is retrieved from the rule library; After the original data collection is completed, it is stored in the original database.

5. The intelligent survey data analysis method based on AI computing power according to claim 4 is characterized in that: Determining an upper limit of the data access amount per unit time corresponding to each data port according to the difference between the data update frequency deviation rate and the data verification failure rate includes: The maximum permissible data access amount per unit time of the target data port is accessed, wherein the value range of the unit time is 1s-5s; Retrieve the actual retrieved data volume corresponding to each unit time of the target data port from the historical operation parameters; Obtain a reference value of the data access amount per unit time of the target data port according to the actual access data amount corresponding to each unit time and the maximum allowed data access amount; Performing a ratio processing on the data retrieval amount reference value and the maximum allowable data retrieval amount to obtain a data retrieval reference ratio; comparing the difference between the data update frequency deviation rate and the data verification failure rate with a data retrieval reference ratio; When the difference between the data update frequency deviation rate and the data verification failure rate is not less than the data retrieval reference ratio, the data retrieval amount per unit time reference value of the target data port is used as the upper limit value of the data retrieval amount per unit time of the target data port; When the difference between the data update frequency deviation rate and the data verification failure rate is less than the data retrieval reference ratio, the deviation failure coupling coefficient corresponding to the target data port is retrieved, where the value range of the deviation failure coupling coefficient is [0.4, 0.8]; The upper limit of the data retrieval amount per unit time of the target data port is set by using the deviation failure coupling coefficient combined with the data update frequency deviation rate and the data verification failure rate.

6. The intelligent survey data analysis method based on AI computing power according to claim 5 is characterized in that: The constructed knowledge graph is intelligently analyzed for: Retrieve sentiment analysis models, cluster analysis models, and regression prediction models from the model library; The sentiment analysis model is to use the sentiment analysis model to perform semantic understanding and sentiment judgment on the text data in the constructed knowledge graph; The clustering analysis model is as follows: the entity attribute data in the constructed knowledge graph is imported into the clustering analysis model. The clustering analysis model automatically aggregates similar entities into different categories based on the set algorithm and parameters; The regression prediction model is as follows: historical data and related attribute features in the constructed knowledge graph are input into the regression prediction model. The regression prediction model predicts future trends by learning the patterns in the data. Integrate the analysis results of each model and establish the correlation between the analysis results; Finally, intelligent analysis data is obtained.

7. The intelligent survey data analysis method based on AI computing power according to claim 6 is characterized in that: Perform dynamic early warning processing based on intelligent analysis results, used for: According to the objectives and fields of survey data analysis, early warning indicators are screened from intelligent analysis data; Among them, early warning indicators include product sales growth rate, proportion of dissatisfied consumers and changes in competitor market share; And determine the specific location of each warning indicator data in the intelligent analysis results; Retrieve historical warning data from the database, determine the fluctuation range and change pattern of each warning indicator based on the historical warning data, and set warning thresholds based on the fluctuation range and change pattern; Among them, the warning thresholds include high risk, medium risk and low risk; Compare and analyze intelligent analysis data with the set warning threshold through automated procedures; Determine whether the intelligent analysis data reaches or exceeds the set warning threshold based on the comparative analysis results; Finally, an early warning response is made based on the judgment results, and the early warning response is the result of dynamic early warning processing.

8. The intelligent survey data analysis method based on AI computing power according to claim 7 is characterized in that: Dynamically adjust the set warning thresholds based on the fluctuation range, including: Determine the fluctuation range of each warning indicator based on historical warning data; Extracting a maximum value and a minimum value of a fluctuation range from the fluctuation range of the early warning indicator; Performing ratio processing on the minimum value of the fluctuation amplitude and the maximum value of the fluctuation amplitude to obtain a fluctuation amplitude ratio parameter; Extract the data port obtained from the early warning indicators; When the target data port does not exist in the data port obtained by the early warning indicator, the set early warning threshold is not adjusted; When a target data port exists in the data ports obtained by the early warning indicator, the data update frequency deviation rate and the deviation failure coupling coefficient corresponding to the target data port are retrieved; The data update frequency deviation rate corresponding to the target data port in the data port obtained by the early warning indicator is combined with the deviation failure coupling coefficient to adjust the set early warning threshold.

9. An intelligent survey data analysis system based on AI computing power, applied to an intelligent survey data analysis method based on AI computing power as claimed in claim 8, characterized in that: include: Data visualization unit, used for: Visualize intelligent analysis results and early warning response data; Visualization processing involves confirming the display data and warning corresponding data in the intelligent analysis results. The display data in the intelligent analysis results include trend data, proportion data, relationship data, and comparison data. Warning response data includes warning indicator status, warning level distribution and response processing progress; Use visualization tools to visualize the display data, and obtain chart data and text data after the visualization is completed; Evaluate the effect of the completed visualization data. The evaluation includes: evaluating the clarity, aesthetics, and accuracy of the visualization data, and optimizing it based on the evaluation results. After the optimization is completed, the visual data of the survey data analysis is obtained; Report generation transmission unit for: Generate visual reports using report generation software using the visual data from survey data analysis; Confirming the report transmission volume of the generated visualization report, wherein the report transmission volume is confirmed according to the size of the visualization report; Confirm the remaining capacity of the transmission channels of each data receiving port, where the number of transmission channels of each data receiving port is not less than 4; Selecting a transmission channel whose remaining capacity is greater than the reported transmission amount as the transmission channel for visualizing the report and transmitting it to different data receiving ports; Finally, the staff receives and views the visualization report at different data receiving ports.

Images