Time sequence database information storage and management system

Through the timing database information storage and management system, performance bottlenecks and security risks of large-scale timing data applications are solved, efficient data processing and security management are realized, and business decision-making and operation optimization are supported.

CN119003613AActive Publication Date: 2024-11-22SHENZHEN TI FENG INTELLIGENT CO LTD
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202411036469.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-22
Estimated Expiration
2044-07-31

AI Technical Summary

Technical Problem

The existing technology has performance bottlenecks when facing large-scale time series data applications, which are difficult to scale, and storage devices are vulnerable to hackers, data management is complex and security risks are high, resulting in difficulty in data sharing and opening up.

Method used

Design a time series database information storage and management system, adopting modular design and advanced data processing technology, including data reception, storage, query, management and analysis modules, supports multi-protocol data reception, distributed storage, user permission management and system monitoring, through The prediction model is used for in-depth analysis.

Benefits of technology

It realizes efficient data reception, storage and query, ensures data security and reliability, supports business decision-making and operational optimization, and improves data processing performance and accuracy through distributed storage and prediction models.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119003613A_ABST
    Figure CN119003613A_ABST
Patent Text Reader

Abstract

The invention discloses a time sequence database information storage and management system, which comprises a data receiving module used for receiving time sequence data from different data sources and converting the time sequence data into a uniform format; the data storage module is used for storing the converted time sequence data in a database and sorting the converted time sequence data according to a time sequence; the data query module is used for retrieving corresponding time sequence data from the database according to a query request of a user and displaying the time sequence data to the user in a chart or table form; the data management module is used for performing addition, deletion, modification and check operations on the time series data in the database; and the data analysis processing module is used for carrying out preprocessing operations such as cleaning, duplicate removal and missing value processing on the received time sequence data. Through the modular design and the advanced data processing technology, a powerful, flexible and efficient tool is provided for an enterprise and used for managing, analyzing and predicting time sequence data, and therefore better service decision making and operation optimization are supported.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of information storage, and more specifically, particularly relates to a time series database information storage and management system. Background Art

[0002] Time series data is data recorded in chronological order, usually containing timestamps, measurement values, and related tags or metadata. This data is generated in large quantities in the fields of IoT devices, financial stock markets, environmental monitoring, etc. Time series data has some significant characteristics: each data point has a timestamp, which is very important for data analysis and calculation; this type of data is usually structured, with predefined data types or fixed lengths;

[0003] Existing technologies often show performance bottlenecks when facing large-scale data applications, such as cloud computing, big data, and artificial intelligence, and are difficult to effectively expand to cope with the growing data demand. Many enterprises have to frequently replace storage devices due to the non-scalability of the system, which increases costs and operational complexity;

[0004] At the same time, storage devices connected to the Internet are vulnerable to hacker attacks, leading to data leaks. Common storage methods such as mobile phones, personal computers, servers and cloud storage all have security risks. The information security industry mainly focuses on information transmission protection and attack defense, but ignores storage security. For example, the authenticity, confidentiality and non-repudiation of stored information are rarely addressed, resulting in major security risks in the field of information storage;

[0005] In addition, due to the different needs of different entities for informatization, various types of data are sealed in different systems, forming data islands, which is not conducive to data sharing and openness. Different units choose different storage formats and database designs, which makes data management and integration extremely complicated and increases management costs. Therefore, we propose a time series database information storage and management system. Summary of the invention

[0006] The purpose of the present invention is to solve the shortcomings of the prior art, and the proposed time series database information storage and management system, through its modular design and advanced data processing technology, provides enterprises with a powerful, flexible and efficient tool for managing, analyzing and predicting time series data, thereby supporting better business decisions and operational optimization.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] Time series database information storage and management system, including:

[0009] The data receiving module is used to receive time series data from different data sources and convert them into a unified format;

[0010] The data storage module is used to store the converted time series data in the database and sort them according to the time series;

[0011] The data query module is used to retrieve the corresponding time series data from the database according to the user's query request and display it to the user in the form of a chart or table;

[0012] The data management module is used to perform operations such as adding, deleting, modifying, and querying time series data in the database, as well as backup and recovery operations;

[0013] The data analysis and processing module is used to perform preprocessing operations such as cleaning, deduplication and missing value processing on the received time series data, as well as statistical analysis, trend prediction and anomaly detection on the stored time series data.

[0014] Preferably, the system further comprises:

[0015] User authority management module, used to control and manage user operation authority;

[0016] The system monitoring module is used to monitor and alarm the system's operating status in real time.

[0017] Preferably, the data receiving module includes:

[0018] Multi-protocol support submodule, which receives data through multiple communication protocols and can parse and unify these data streams from different protocols;

[0019] Data format parser, used to load different parsing plug-ins to parse the received data;

[0020] Data transformation and mapping engine, which maps raw data into the database schema, including processing times and values ​​for time series data, and mapping of any metadata tags;

[0021] The data quality monitoring unit performs quality checks on the data entering the system in real time. The checks include but are not limited to range checks, integrity checks, and accuracy tests to ensure that only high-quality data is accepted and stored.

[0022] Preferably, the data storage module adopts a distributed storage architecture, and stores data in multiple nodes in a dispersed manner through sharding technology to achieve horizontal expansion, and adopts an optimized writing strategy to obtain higher writing performance through memory writing and subsequent disk sequential writing, thereby avoiding the overhead of random writing;

[0023] The data storage module should have a built-in data compression and downsampling unit, utilize the regularity of time series data, adopt differential compression, dictionary compression and other algorithms to reduce data volume, and automatically downsample to merge high-precision data into low-precision data to further save storage space.

[0024] Preferably, the specific workflow of the data query module is:

[0025] 1) Users construct queries through the query interface and specify time range, dimension, and aggregation function conditions;

[0026] 2) The query module parses the query request submitted by the user and passes it to the query optimizer to generate an efficient execution plan;

[0027] 3) According to the optimized execution plan, the query engine accesses the relevant data indexes and storage and executes the query operation;

[0028] 4) Return the query result set to the user, support paging and on-demand loading, and improve the performance of large data processing;

[0029] 5) Display the data in charts or other forms.

[0030] Preferably, the data analysis and processing module establishes a prediction model to analyze and process the data, and the specific process is as follows:

[0031] 1) Identify the business problems or analysis objectives that need to be solved and develop a data analysis plan for the problem;

[0032] 2) Determine the scope and type of data required based on analysis requirements and obtain data from designated sources;

[0033] 3) Build applicable prediction models according to the analysis purpose and train the models;

[0034] 4) Present the analysis results in the form of reports, charts, and visual interfaces, adjust the analysis model and process based on business feedback, continuously monitor the analysis results, and perform iterative optimization.

[0035] Preferably, the prediction model is selected according to the importance and relevance of the features to reduce the model complexity and avoid overfitting, and the partial derivative of the loss function with respect to the weight is used to guide the update of the weight, and the expression is:

[0036]

[0037] Where η is the learning rate, is the partial derivative of the loss function with respect to the current weight, E is the loss function, and w is any weight in the network;

[0038] The prediction model is trained using the back propagation algorithm. First, the error of the output layer is calculated, and then it is back propagated layer by layer to the input layer. For each layer, the algorithm calculates the contribution of each neuron to the total error and adjusts the weight accordingly. If the input of neuron j is net j , the output is o j , the error is δ j , with weight w ji , the expression is:

[0039] Δw ji =η·δ j o i ;

[0040] In the formula, δ j It is calculated based on the type of neuron and the activation function.

[0041] Preferably, the prediction model uses a nonlinear function during training so that the model can capture complex data features, and the expression is:

[0042]

[0043] In the formula, σ(x) represents the output of the Sigmoid function, which is a real number with a value range between 0 and 1, x is the real value input to the Sigmoid function, e is the base of the natural logarithm, -x represents the opposite of the input value, and 1+e -x It is the denominator part that ensures that the result of the fraction is mapped between 0 and 1.

[0044] Preferably, in the prediction model, a loss function is used to measure the difference between the predicted value and the actual value of the neural network, and the expression is:

[0045]

[0046] Where M is the number of categories, y c is the true label, is the predicted output, log is the logarithmic function with the natural constant e as the base;

[0047] The calculation process of the above formula is:

[0048] For each category c, calculate the true label yc and the predicted probability The product of

[0049] Take the logarithm of this product

[0050] Adding all these values ​​together gives us the final loss E.

[0051] Compared with the prior art, the time series database information storage and management system provided by the present invention has the following advantages:

[0052] Efficient data reception and conversion: By supporting multi-protocol data reception modules, the system can process time series data from different data sources and convert them into a unified format; this improves data availability and consistency, and provides a solid foundation for subsequent data analysis and management;

[0053] Optimized data storage: With distributed storage architecture and sharding technology, the system can achieve horizontal expansion and improve data writing performance; built-in data compression and downsampling units further save storage space while ensuring data integrity and accuracy;

[0054] Flexible data query and display: The data query module allows users to retrieve data based on different time ranges, dimensions, and aggregation function conditions; query results can be displayed in the form of charts or tables, supporting paging and on-demand loading, improving the performance of large data processing;

[0055] Comprehensive data management: The data management module not only supports the addition, deletion, modification and query of time series data, but also includes backup and recovery functions to ensure data security and reliability;

[0056] In-depth data analysis and processing: The data analysis and processing module can perform pre-processing operations such as cleaning, deduplication, and missing value processing on time series data, and perform analytical operations such as statistical analysis, trend prediction, and anomaly detection; this helps to extract valuable information from the data and support decision-making;

[0057] User rights management and system monitoring: The user rights management module ensures the security of the system and prevents unauthorized access and operation; the system monitoring module monitors the operating status of the system in real time, and promptly detects and alerts potential problems;

[0058] Intelligent prediction model: By building and training the prediction model, the system can conduct in-depth analysis of time series data, provide business insights and predict future trends. The model training uses back propagation algorithms and nonlinear functions to capture complex data features and improve prediction accuracy.

[0059] Application of loss function: The prediction model uses the loss function to measure the difference between the predicted value and the actual value of the neural network. By optimizing the loss function, the model weight is adjusted to improve the performance of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a diagram of the architecture of the time series database information storage and management system of the present invention;

[0061] Figure 2The following is a specific workflow diagram of the data query module in an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0063] The present invention provides a time series database information storage and management system, which, through its modular design and advanced data processing technology, provides enterprises with a powerful, flexible and efficient tool for managing, analyzing and predicting time series data, thereby supporting better business decisions and operational optimization.

[0064] System includes:Includes:

[0065] The data receiving module is used to receive time series data from different data sources and convert it into a unified format; the data receiving module includes:

[0066] Multi-protocol support submodule, which receives data through multiple communication protocols and can parse and unify these data streams from different protocols;

[0067] Data format parser, used to load different parsing plug-ins to parse the received data;

[0068] Data transformation and mapping engine, which maps raw data into the database schema, including processing times and values ​​for time series data, and mapping of any metadata tags;

[0069] The data quality monitoring unit performs quality checks on the data entering the system in real time, including but not limited to range checks, integrity checks, and accuracy tests, to ensure that only high-quality data is accepted and stored;

[0070] The data storage module is used to store the converted time series data in the database and sort them according to the time series; the data storage module adopts a distributed storage architecture, and stores the data in multiple nodes through sharding technology to achieve horizontal expansion, and adopts an optimized writing strategy to obtain higher writing performance through memory writing and subsequent disk sequential writing, avoiding the overhead of random writing;

[0071] The data storage module should have a built-in data compression and downsampling unit, utilize the regularity of time series data, adopt differential compression, dictionary compression and other algorithms to reduce data volume, and automatically downsample to merge high-precision data into low-precision data to further save storage space.

[0072] The data query module is used to retrieve the corresponding time series data from the database according to the user's query request and display it to the user in the form of a chart or table; Figure 2 , the specific workflow of the data query module is as follows:

[0073] 1) Users construct queries through the query interface and specify time range, dimension, and aggregation function conditions;

[0074] 2) The query module parses the query request submitted by the user and passes it to the query optimizer to generate an efficient execution plan;

[0075] 3) According to the optimized execution plan, the query engine accesses the relevant data indexes and storage and executes the query operation;

[0076] 4) Return the query result set to the user, support paging and on-demand loading, and improve the performance of large data processing;

[0077] 5) Display the data in charts or other forms.

[0078] The data management module is used to perform operations such as adding, deleting, modifying, and querying time series data in the database, as well as backup and recovery operations;

[0079] Specifically, the functions of the data management module are as follows:

[0080] Data addition, deletion, modification and query operations:

[0081] Adding data allows users to add new time series data to the database; this usually involves validating and formatting the data to ensure that the structure and integrity of the data meet system requirements;

[0082] Delete data. Users can delete time series data that is no longer needed from the database. The deletion operation needs to be handled with caution to avoid accidentally deleting important data.

[0083] To modify data, users can update existing time series data in the database. The modification operation needs to ensure data consistency and verify the validity of the new data.

[0084] Query data. Users can query time series data based on specific conditions, such as time range, specific indicators, etc.; query operations need to be optimized to support fast data retrieval.

[0085] Data backup:

[0086] Automatic backup, the system can be configured to automatically back up data regularly to prevent data loss;

[0087] Manual backup: users can also manually trigger the data backup process as needed;

[0088] Incremental backup and full backup, support incremental backup and full backup;

[0089] Data Recovery

[0090] Disaster recovery: in case of data damage or loss, backup data can be used for recovery;

[0091] Restore to a specific point: supports restoring data to a specific point in time, which is very useful for recovering data lost due to operational errors.

[0092] Data security and rights management

[0093] Access control ensures that only authorized users can perform operations such as adding, deleting, modifying, and querying data;

[0094] Data encryption: encrypt sensitive data to protect the security of data during storage and transmission.

[0095] Through the above functions and technologies, the data management module can effectively protect and manage time series data, ensure data availability and security, and provide flexible data operation capabilities.

[0096] The data analysis and processing module is used to perform preprocessing operations such as cleaning, deduplication and missing value processing on the received time series data, as well as statistical analysis, trend prediction and anomaly detection on the stored time series data; the data analysis and processing module establishes a prediction model to analyze and process the data. The specific process is as follows:

[0097] 1) Identify the business problems or analysis objectives that need to be solved and develop a data analysis plan for the problem;

[0098] 2) Determine the scope and type of data required based on analysis requirements and obtain data from designated sources;

[0099] 3) Build applicable prediction models according to the analysis purpose and train the models;

[0100] 4) Present the analysis results in the form of reports, charts, and visual interfaces, adjust the analysis model and process based on business feedback, continuously monitor the analysis results, and perform iterative optimization;

[0101] The prediction model is selected according to the importance and relevance of the features to reduce the model complexity and avoid overfitting. The partial derivative of the loss function with respect to the weight is used to guide the update of the weight. The expression is:

[0102]

[0103] Where η is the learning rate, is the partial derivative of the loss function with respect to the current weight, E is the loss function, and w is any weight in the network;

[0104] The prediction model is trained using the back propagation algorithm. First, the error of the output layer is calculated, and then it is back propagated layer by layer to the input layer. For each layer, the algorithm calculates the contribution of each neuron to the total error and adjusts the weight accordingly. If the input of neuron j is net j , the output is o j , the error is δ j , with weight w ji , the expression is:

[0105] Δw ji =η·δ j o i ;

[0106] In the formula, δ j It is calculated based on the type of neuron and the activation function;

[0107] The prediction model uses a nonlinear function during training to enable the model to capture complex data features. The expression is:

[0108]

[0109] In the formula, σ(x) represents the output of the Sigmoid function, which is a real number with a value range between 0 and 1, x is the real value input to the Sigmoid function, e is the base of the natural logarithm, -x represents the opposite of the input value, and 1+e -x It is the denominator part, which ensures that the result of the fraction will be mapped between 0 and 1;

[0110] In the prediction model, a loss function is used to measure the difference between the predicted value and the actual value of the neural network, and the expression is:

[0111]

[0112] Where M is the number of categories, y c is the true label, is the predicted output, log is the logarithmic function with the natural constant e as the base;

[0113] The calculation process of the above formula is:

[0114] For each category c, calculate the true label y c and predicted probability The product of

[0115] Take the logarithm of this product

[0116] Adding all these values ​​together gives us the final loss E.

[0117] User authority management module, used to control and manage user operation authority;

[0118] Specifically, the functions of the user rights management module are as follows:

[0119] User Authentication:

[0120] Login verification, which verifies the credentials provided by the user (such as username and password) to confirm their identity;

[0121] Multi-factor authentication: support for multi-factor authentication (MFA) to increase the security layer, such as combining passwords and mobile phone verification codes;

[0122] Role-based access control:

[0123] Role definition, defining different user roles, each with a specific set of permissions;

[0124] Role allocation: assign roles to specific users or user groups to simplify permission management;

[0125] 3. Permission management:

[0126] Fine-grained permission control can accurately control the data and functions that users can access and operate;

[0127] Data-level permissions restrict users to access only specific data sets and prevent them from accessing other users’ data;

[0128] Audit and Compliance:

[0129] Operation logs record all user operations, including login, data query, modification, etc., for auditing and backtracking;

[0130] Compliance reporting, automatically generates compliance reports that show detailed records of user activities to meet regulatory requirements.

[0131] The user rights management module provides a secure and controllable access control environment for the time series database information storage and management system, ensuring that only authorized users can access and operate data in the system, while providing the necessary audit and compliance support to meet the enterprise's security and regulatory requirements.

[0132] System monitoring module, used for real-time monitoring and alarm of the system's operating status;

[0133] Specifically, the functions of the system monitoring module are as follows:

[0134] Performance Monitoring:

[0135] Response time monitoring, tracking the response time of the system processing requests to ensure that the service responds quickly;

[0136] Throughput monitoring measures the number of requests processed by the system per unit time and evaluates system capabilities;

[0137] Error Monitoring:

[0138] Error rate monitoring, tracking the error rate in the system, such as database query failure rate or API error rate;

[0139] Exception capture: capture exceptions in the system and record detailed information for subsequent analysis;

[0140] Resource usage monitoring:

[0141] CPU usage, monitor the system's CPU usage to ensure that the processor is not overused;

[0142] Memory usage, track memory usage and prevent memory leaks or shortages;

[0143] Disk space monitoring: monitor disk usage to ensure there is enough storage space;

[0144] Network Monitoring:

[0145] Network traffic monitoring: monitor the traffic of the network interface to ensure that the network connection is stable and the bandwidth is sufficient;

[0146] Connection monitoring, monitoring the connection status between the database and the application;

[0147] Alarm mechanism:

[0148] Threshold setting: set thresholds for various monitoring indicators, and trigger alarms once the thresholds are exceeded;

[0149] Notification method: sending alarm notifications via email, SMS or instant messaging;

[0150] The system monitoring module can ensure the efficient operation of the time series database information storage and management system, and promptly discover and resolve potential problems, thereby ensuring the stability and availability of the system.

[0151] In summary, compared with the prior art, the present invention has the following advantages:

[0152] Efficient data reception and conversion: By supporting multi-protocol data reception modules, the system can process time series data from different data sources and convert them into a unified format; this improves data availability and consistency, and provides a solid foundation for subsequent data analysis and management;

[0153] Optimized data storage: With distributed storage architecture and sharding technology, the system can achieve horizontal expansion and improve data writing performance; built-in data compression and downsampling units further save storage space while ensuring data integrity and accuracy;

[0154] Flexible data query and display: The data query module allows users to retrieve data based on different time ranges, dimensions, and aggregation function conditions; query results can be displayed in the form of charts or tables, supporting paging and on-demand loading, improving the performance of large data processing;

[0155] Comprehensive data management: The data management module not only supports the addition, deletion, modification and query of time series data, but also includes backup and recovery functions to ensure data security and reliability;

[0156] In-depth data analysis and processing: The data analysis and processing module can perform pre-processing operations such as cleaning, deduplication, and missing value processing on time series data, and perform analytical operations such as statistical analysis, trend prediction, and anomaly detection; this helps to extract valuable information from the data and support decision-making;

[0157] User rights management and system monitoring: The user rights management module ensures the security of the system and prevents unauthorized access and operation; the system monitoring module monitors the operating status of the system in real time, and promptly detects and alerts potential problems;

[0158] Intelligent prediction model: By building and training the prediction model, the system can conduct in-depth analysis of time series data, provide business insights and predict future trends. The model training uses back propagation algorithms and nonlinear functions to capture complex data features and improve prediction accuracy.

[0159] Application of loss function: The prediction model uses the loss function to measure the difference between the predicted value and the actual value of the neural network. By optimizing the loss function, the model weight is adjusted to improve the performance of the model.

[0160] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Time series database information storage and management system, It is characterized in that include: The data receiving module is used to receive time series data from different data sources and convert them into a unified format; The data storage module is used to store the converted time series data in the database and sort them according to the time series; The data query module is used to retrieve the corresponding time series data from the database according to the user's query request and display it to the user in the form of a chart or table; The data management module is used to perform operations such as adding, deleting, modifying, and querying time series data in the database, as well as backup and recovery operations; The data analysis and processing module is used to perform preprocessing operations such as cleaning, deduplication and missing value processing on the received time series data, as well as statistical analysis, trend prediction and anomaly detection on the stored time series data.

2. According to claim 1, the time series database information storage and management system, It is characterized in that The system also includes: User authority management module, used to control and manage user operation authority; The system monitoring module is used to monitor and alarm the system's operating status in real time.

3. According to claim 1, the time series database information storage and management system, It is characterized in that The data receiving module comprises: Multi-protocol support submodule, which receives data through multiple communication protocols and can parse and unify these data streams from different protocols; Data format parser, used to load different parsing plug-ins to parse the received data; Data transformation and mapping engine, which maps raw data into the database schema, including processing times and values ​​for time series data, and mapping of any metadata tags; The data quality monitoring unit performs quality checks on the data entering the system in real time. The checks include but are not limited to range checks, integrity checks, and accuracy tests to ensure that only high-quality data is accepted and stored.

4. The time series database information storage and management system according to claim 1, It is characterized in that The data storage module adopts a distributed storage architecture, and stores data in multiple nodes through sharding technology to achieve horizontal expansion. It also adopts an optimized writing strategy to obtain higher writing performance through memory writing and subsequent disk sequential writing, thus avoiding the overhead of random writing. The data storage module should have a built-in data compression and downsampling unit, utilize the regularity of time series data, adopt differential compression, dictionary compression and other algorithms to reduce data volume, and automatically downsample to merge high-precision data into low-precision data to further save storage space.

5. The time series database information storage and management system according to claim 1, It is characterized in that The specific workflow of the data query module is as follows: 1) Users construct queries through the query interface and specify time range, dimension, and aggregation function conditions; 2) The query module parses the query request submitted by the user and passes it to the query optimizer to generate an efficient execution plan; 3) According to the optimized execution plan, the query engine accesses the relevant data indexes and storage and executes the query operation; 4) Return the query result set to the user, support paging and on-demand loading, and improve the performance of large data processing; 5) Display the data in charts or other forms.

6. The time series database information storage and management system according to claim 1, It is characterized in that The data analysis and processing module establishes a prediction model to analyze and process the data. The specific process is as follows: 1) Identify the business problems or analysis objectives that need to be solved and develop a data analysis plan for the problem; 2) Determine the scope and type of data required based on analysis requirements and obtain data from designated sources; 3) Build applicable prediction models according to the analysis purpose and train the models; 4) Present the analysis results in the form of reports, charts, and visual interfaces, adjust the analysis model and process based on business feedback, continuously monitor the analysis results, and perform iterative optimization.

7. The time series database information storage and management system according to claim 6, It is characterized in that The prediction model is selected according to the importance and relevance of the features to reduce the model complexity and avoid overfitting. The partial derivative of the loss function with respect to the weight is used to guide the update of the weight. The expression is: Where η is the learning rate, is the partial derivative of the loss function with respect to the current weight, E is the loss function, and w is any weight in the network; The prediction model is trained using the back propagation algorithm. First, the error of the output layer is calculated, and then it is back propagated layer by layer to the input layer. For each layer, the algorithm calculates the contribution of each neuron to the total error and adjusts the weight accordingly. If the input of neuron j is net j , the output is o j , the error is δ j , with weight w ji , the expression is: Δw ji =η·δ j ·o i ; In the formula, δ j It is calculated based on the type of neuron and the activation function.

8. The time series database information storage and management system according to claim 6, It is characterized in that The prediction model uses a nonlinear function during training to enable the model to capture complex data features. The expression is: In the formula, σ(x) represents the output of the Sigmoid function, which is a real number with a value range between 0 and 1, x is the real value input to the Sigmoid function, e is the base of the natural logarithm, -x represents the opposite of the input value, and 1+e -x It is the denominator part that ensures that the result of the fraction is mapped between 0 and 1.

9. The time series database information storage and management system according to claim 8, It is characterized in that In the prediction model, a loss function is used to measure the difference between the predicted value and the actual value of the neural network, and the expression is: Where M is the number of categories, y c is the true label, is the predicted output, log is the logarithmic function with the natural constant e as the base; The calculation process of the above formula is: For each category c, calculate the true label y c and predicted probability The product of Take the logarithm of this product Adding all these values ​​together gives us the final loss E.

Citation Information

Patent Citations

  • Load prediction method based on data enhancement strategy and residual error self-correction learning model

    CN115936176A

  • Intelligent operation and maintenance data storage platform

    CN117235169A

  • Basic graduation destination prediction method based on causal inference and visual analysis system

    CN117609731A

  • Hudi asynchronous data clustering method and system based on hotspot prediction

    CN117633232A

  • Efficient storage system for big data statistical collection

    CN117971129A