Database information prediction method and computer device

Abstract

This application provides a database information prediction method and computer device, belonging to the field of database technology. The method includes: determining multi-dimensional characteristics of the target database based on its capacity metadata and performance index data; obtaining prediction results output by each prediction model based on the multi-dimensional characteristics of the target database and multiple pre-built prediction models; and determining a target prediction result for the target database based on the prediction results output by each prediction model and the multi-dimensional characteristics of the target database. The target prediction result includes the number of days the target database's capacity is available and a confidence interval. This application can adapt to the non-linear growth and periodic fluctuations of the database and adaptively adjust the prediction results based on the database's business data, thereby improving the accuracy of database capacity prediction.

Description

Technical Field

[0001] This application relates to the field of database technology, and more specifically, to a database information prediction method and a computer device. Background Technology

[0002] With the increasing informatization of enterprises and the explosive growth of business data, the effective management and planning of the storage capacity of relational databases, as the core data carrier, has become crucial to ensuring business continuity and stability. However, traditional database capacity management methods mainly rely on manual monitoring and experience-based judgment by operations and maintenance personnel, which suffers from many problems such as low efficiency, delayed response, and inaccurate prediction.

[0003] In related technologies, linear prediction models are often used to predict database capacity. Specifically, by calculating the average daily increase in database capacity over a period of time, the time point at which the database capacity will reach its limit is derived, and alarm thresholds are set or expansion processes are triggered accordingly.

[0004] However, when using related technologies to predict database capacity, there are problems such as inability to adapt to nonlinear growth and periodic fluctuations, lack of effective identification and processing mechanisms for abnormal data, inability to adaptively adjust the prediction model according to the latest business data, single prediction dimension, and lack of a capacity health assessment system. Summary of the Invention

[0005] The purpose of this application is to provide a database information prediction method and computer equipment that can adapt to the nonlinear growth and periodic fluctuations of the database, and adaptively adjust the prediction results based on the business data of the database, thereby improving the accuracy of database capacity prediction.

[0006] The embodiments of this application are implemented as follows: A first aspect of this application provides a database information prediction method, the method comprising: Based on the target database's capacity metadata and performance metrics, determine the target database's multi-dimensional characteristics, including: daily data growth, long-term trend characteristics, periodic characteristics, business-related characteristics, and growth acceleration. Based on the multi-dimensional features of the target database and multiple pre-built prediction models, the prediction results output by each prediction model are obtained. Based on the prediction results output by each prediction model and the multi-dimensional characteristics of the target database, the target prediction results for the target database are determined. The target prediction results include: the number of days of available capacity for the target database and the confidence interval.

[0007] As one possible implementation, based on the target database's capacity metadata and performance metrics, multi-dimensional characteristics of the target database are determined, including: Obtain connection information and data acquisition strategies for the target database; Establish a connection with the target database based on the connection information; According to the data acquisition strategy, the corresponding capacity metadata and performance index data of the target database are collected from the data dictionary and dynamic performance view of the target database, and historical time series data are composed according to the timestamp of the capacity metadata and the timestamp of the performance index data. The historical time series data is preprocessed, and multi-dimensional features of the target database are extracted from the preprocessed historical time series data.

[0008] As one possible implementation, historical time-series data is preprocessed, and multi-dimensional features of the target database are extracted from the preprocessed historical time-series data, including: Historical time-series data is cleaned and standardized to obtain structured time-series data that meets the preset format; Anomaly detection is performed on structured time series data to filter out abnormal data from the structured time series data; Extract multi-dimensional features of the target database from structured time-series data after filtering out outlier data.

[0009] As one possible implementation, outlier detection is performed on structured time-series data to filter out anomalous data, including: Determine the first data point of the structured time series data at the first percentile position, and determine the second data point of the structured time series data at the second percentile position; Based on the first and second data, determine the interquartile range corresponding to the structured time series data; Based on the first data and the interquartile range, determine the lower bound of outlier data corresponding to the structured time series data; Based on the second data and the interquartile range, determine the upper bound of the outlier data corresponding to the structured time series data; Based on the upper and lower bounds of outlier data, outlier data is filtered out from structured time series data.

[0010] As one possible implementation, the aforementioned prediction models include: a basic linear model, a periodic model, and a dynamic weight allocation model; based on the multi-dimensional features of the target database and the pre-built prediction models, the prediction results output by each prediction model are obtained, including: The daily growth rate, growth acceleration, and long-term trend characteristics of the data in the multi-dimensional features are input into the basic linear model to obtain the linear prediction results of the target database. The linear prediction results are used to characterize the short-term data change trend of the target database. Periodic features from multidimensional features are input into a periodic model to obtain periodic prediction results for the target database. The periodic prediction results are used to characterize the periodic data change trend of the target database. The linear and periodic prediction results are input into the dynamic weight allocation model to obtain the target weight sequence, which includes the weight values ​​of the basic linear model and the weight values ​​of the periodic model.

[0011] As one possible implementation, the aforementioned prediction models also include: a business-aware model; and, based on the prediction results output by each prediction model and the multi-dimensional features of the target database, determining the target prediction results for the target database, including: The linear and periodic prediction results are weighted and fused based on the target weight series to obtain the initial prediction results for the target database. The business-related features from the multi-dimensional features and the initial prediction results of the target database are input into the business-aware model to obtain the target prediction results of the target database.

[0012] As one possible implementation, business-related features from the multi-dimensional features and the initial prediction results of the target database are input into the business-aware model to obtain the target prediction results for the target database, including: Determine the sequence of factors influencing data changes in the target database within the future time period. The sequence of factors influencing data changes includes: factors influencing peak business periods, factors influencing statutory holidays, and factors influencing business events to be executed. Based on the business-aware model, the initial prediction results of the target database are adjusted according to the sequence of factors affecting data changes, so as to obtain the target prediction results of the target database.

[0013] As one possible implementation, the above-mentioned database information prediction method also includes: Based on the target prediction results of the target database and the weight parameters of the preset database type, the health score result of the target database is determined; Based on the health score and target prediction results of the target database, tiered early warning or automatic expansion is performed on the target database respectively.

[0014] As one possible implementation, the health score of the target database is determined based on the target prediction results and the weight parameters of the preset database type, including: Based on the number of days of available capacity in the target prediction results, determine the capacity utilization score of the target database; Based on multi-dimensional features, historical capacity growth data of the target database is obtained by fitting, and the growth rate trend score of the target database is determined based on the historical capacity growth data. Based on the database type of the target database, determine the corresponding weight parameters for the target database. The weight parameters are used to indicate the importance of the capacity utilization rate and growth rate trend of the target database in the health score results. Based on the capacity utilization score, the growth rate trend score, and the weighting parameters, the health score of the target database is determined.

[0015] A second aspect of this application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the database information prediction method described in the first aspect.

[0016] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the database information prediction method described in the first aspect.

[0017] The beneficial effects of the embodiments of this application include: This application provides a database information prediction method. It acquires capacity metadata and performance index data of a target database, and constructs historical time-series data of the target database based on the timestamps of the capacity metadata and performance index data. From this historical time-series data, it extracts multi-dimensional features of the target database, including daily growth, long-term trend characteristics, periodic characteristics, business-related characteristics, and growth acceleration. These multi-dimensional features are then input in parallel into multiple pre-built prediction models. Each prediction model performs capacity prediction for one dimension of the target database based on its built-in algorithm and outputs the corresponding prediction result. The prediction results from each model are weighted and fused to obtain an initial prediction result for the target database. The initial prediction result is then adaptively adjusted based on the business-related features of the target database's multi-dimensional features to obtain the target prediction result. The pre-built prediction models include a basic linear model, a periodic model, a dynamic weight allocation model, and a business-aware model, which can predict from multiple data dimensions, improving the accuracy of the prediction results. Furthermore, this application not only retains the basic linear model but also introduces a periodic model and a business-aware model to enable the prediction results for the target database to adapt to its non-linear growth and periodic fluctuations. Moreover, this application adaptively adjusts the initial prediction results for the target database based on business-related characteristics, thereby improving the accuracy of database capacity prediction. In this way, it achieves the effect of adapting to the non-linear growth and periodic fluctuations of the database and adaptively adjusting the prediction results based on the database's business data, thus improving the accuracy of database capacity prediction. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A system architecture diagram of a database information prediction system provided in this application embodiment; Figure 2 A flowchart illustrating the first database information prediction method provided in this application embodiment; Figure 3 A flowchart illustrating the second database information prediction method provided in this application embodiment; Figure 4 A flowchart illustrating the third database information prediction method provided in this application embodiment; Figure 5A flowchart illustrating the fourth database information prediction method provided in this application embodiment; Figure 6 A flowchart illustrating the fifth database information prediction method provided in this application embodiment; Figure 7 A flowchart of the sixth database information prediction method provided in the embodiments of this application; Figure 8 A flowchart of the seventh database information prediction method provided in the embodiments of this application; Figure 9 A flowchart illustrating the eighth database information prediction method provided in this application embodiment; Figure 10 A mapping table for health scores is provided in the embodiments of this application; Figure 11 A flowchart of the ninth database information prediction method provided in the embodiments of this application; Figure 12 A mapping table for capacity utilization scoring is provided for embodiments of this application; Figure 13 A mapping table for growth rate trend scoring is provided for embodiments of this application; Figure 14 A mapping table for a decision matrix provided in an embodiment of this application; Figure 15 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] In the description of this application, it should be noted that the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] Currently, database capacity is often predicted using linear forecasting models. Specifically, this involves calculating the average daily increase in database capacity over a past period to deduce the timeframe when the database will reach its capacity limit, and then setting alarm thresholds or triggering expansion processes accordingly. However, this approach suffers from several drawbacks, including its inability to adapt to the non-linear growth and periodic fluctuations of databases, a lack of effective mechanisms for identifying and handling abnormal data, an inability to adaptively adjust the forecasting model based on the latest business data, a limited forecasting dimension, and the absence of a comprehensive database capacity health assessment system.

[0024] To address this, this application provides a database information prediction method. By acquiring capacity metadata and performance index data of the target database, multi-dimensional features such as daily growth, long-term trend characteristics, periodic characteristics, business correlation characteristics, and growth acceleration of the target database within a preset historical period are extracted. These multi-dimensional features are then input into multiple pre-constructed prediction models to obtain prediction results from each model. These prediction results are then weighted and fused to obtain an initial prediction result for the target database. Finally, the initial prediction result is corrected based on the business correlation characteristics of the target database to obtain a target prediction result. This method adapts to the non-linear growth and periodic fluctuations of the database and adaptively adjusts the prediction results based on the database's business data, thereby improving the accuracy of database capacity prediction.

[0025] The database information prediction method provided in the embodiments of this application will be explained in detail below with reference to the accompanying drawings.

[0026] Figure 1 A system architecture diagram of a database information prediction system provided in this application is shown below. Figure 1 The database information prediction system provided in this application includes a data acquisition and cleaning module and an intelligent analysis and processing module. The data acquisition and cleaning module includes a data acquisition unit, a data storage unit, a data cleaning unit, and a data analysis unit. The intelligent analysis and processing module includes a prediction engine, a decision engine, an execution engine, and a data storage engine. The prediction engine includes prediction functions such as database capacity prediction, expansion prediction, and health prediction. The decision engine includes decision mechanisms such as capacity expansion mechanisms, data cleaning mechanisms, and data compression mechanisms. The execution engine includes execution methods such as expansion execution, data migration, and data cleaning. The data storage engine includes data management standards such as data classification standards and compressed storage standards.

[0027] The data acquisition and cleaning module is mainly responsible for interacting with the target database to acquire, persist, and process the raw data, providing a high-quality and standardized data foundation for the intelligent analysis and processing module. The intelligent analysis and processing module receives the standardized data and features processed by the data acquisition and cleaning module, realizing an intelligent closed loop from prediction and decision-making to execution.

[0028] Optionally, the data acquisition and cleaning module includes: a data acquisition unit, a data storage unit, a data cleaning unit, and a data analysis unit. The data acquisition unit automatically connects to one or more target database instances based on pre-configured connection information and collects capacity metadata such as tablespace utilization, data file size, table size, index size, and row count, as well as performance metrics such as central controller, memory, and input / output rates, from the target database's data dictionary and dynamic performance view based on a predefined data acquisition strategy for the target database. The data acquisition unit sends the collected timestamped raw data to the data storage unit. The data storage unit persistently stores the received timestamped raw data in the form of time-series data, constructs a historical data sequence of capacity and performance, and sends it to the data cleaning unit. The data cleaning unit performs data preprocessing on the received historical data sequence, including but not limited to format standardization, outlier detection, and smoothing, to eliminate interference from sudden fluctuations such as one-time data imports on subsequent intelligent analysis. The data analysis unit extracts key features from the cleaned historical data sequence, such as calculating daily growth, weekly growth rate, monthly growth rate, and growth acceleration, and identifies periodic fluctuation patterns and correlations with business metrics, forming a multi-dimensional feature set for prediction.

[0029] Optionally, the intelligent analysis and processing module includes: a prediction engine, a decision engine, an execution engine, and a data storage engine. The prediction engine integrates various machine learning models and statistical algorithms to perform prediction functions such as database capacity prediction, expansion prediction, and health prediction. Database capacity prediction is based on the multi-dimensional characteristics of the target database, fusing basic linear models, periodic models, etc., and using a dynamic weight allocation model for weighting to predict the capacity usage trend and available days of the target database over a future period. Expansion prediction is based on the capacity prediction results to determine whether a preset expansion threshold has been reached. Health prediction is based on the capacity prediction results and historical growth trends, combined with configurable weight parameters, to calculate the database's comprehensive health score and determine its risk level.

[0030] In addition, the decision engine determines the optimal operation and maintenance decisions based on the output of the prediction engine (such as the number of days of capacity availability and health score of the target database) and a pre-defined business rule base. The decision engine includes mechanisms such as capacity expansion, data cleaning, and data compression. The capacity expansion mechanism triggers different levels of alerts or automatic expansion decisions based on the number of days of capacity availability and health score of the target database. The data cleaning mechanism identifies archiveable or cleanable historical data, providing suggestions for capacity optimization. The data compression mechanism evaluates and triggers compression operations on non-hotspot data to improve storage efficiency.

[0031] Furthermore, the execution engine is primarily responsible for automating the execution of instructions generated by the decision engine. The execution engine includes execution methods such as scaling, data migration, and data cleanup. Scaling refers to the execution engine automatically calling database commands or application programming interfaces (APIs) to perform scaling operations on tablespaces or data files. Data migration refers to the execution engine migrating data to other storage or archiving systems based on instructions output by the decision engine. Data cleanup refers to the execution engine performing secure data deletion or archiving operations.

[0032] Furthermore, the data storage engine provides internal data management standards for the intelligent analysis and processing module. These standards include data classification standards and compressed storage standards. Specifically, the data classification standards define the classification and storage specifications for intermediate analysis data, model parameters, decision logs, etc., while the compressed storage standards define standard strategies for compressing and storing non-real-time data such as historical analysis results and logs.

[0033] Figure 2 A flowchart of a database information prediction method provided in this application is provided, which is applied to the above-mentioned method. Figure 1 The database information prediction system shown. See also Figure 2 The database information prediction method provided in this application includes: S201. Based on the target database's capacity metadata and performance metrics, determine the target database's multi-dimensional characteristics, including: daily data growth, long-term trend characteristics, periodic characteristics, business-related characteristics, and growth acceleration.

[0034] Among them, the capacity metadata comes from the data dictionary of the target database, which is static data describing the storage structure and current space occupancy status of the target database; the performance index data comes from the dynamic performance view of the target database, which is dynamic data describing the runtime behavior, efficiency and resource consumption of the target database, and is used to reflect the health status and workload of the target database.

[0035] Specifically, the target database's capacity metadata includes: tablespace utilization, data file size, data file growth, table size, index size, partition size, number of rows, and statistical information time, etc. The target database's performance metrics include: resource utilization, load and efficiency, cache efficiency, etc.

[0036] It should be noted that capacity metadata changes relatively slowly, usually only when data is added, deleted, or modified; performance metrics data, on the other hand, is a continuous time-series stream that can change every second.

[0037] Optionally, based on the capacity metadata and performance index data periodically collected from the target database, a deep analysis is performed on the preprocessed historical time-series data to determine multi-dimensional features used to characterize the dynamic changes in database capacity.

[0038] Specifically, the multi-dimensional features include at least: daily data growth, long-term trend characteristics, periodic characteristics, business-related characteristics, and growth acceleration. Daily data growth reflects the daily changes in the target database capacity; long-term trend characteristics are monthly or longer-term trends extracted using time series decomposition methods; periodic characteristics refer to regular fluctuations on a weekly or monthly basis; business-related characteristics are key business features associated with database capacity changes (such as order volume, number of active users, etc.); and growth acceleration is the second derivative characteristic calculated based on changes in daily data growth, used to assess whether the database's data growth trend is accelerating or slowing down.

[0039] S202. Based on the multi-dimensional features of the target database and multiple pre-built prediction models, obtain the prediction results output by each prediction model.

[0040] Optionally, the pre-built prediction models include: a basic linear model, a periodic model, a dynamic weight allocation model, and a business-aware model. The basic linear model captures the stable linear trend of historical time-series data in the target database; the periodic model fits the periodic fluctuation patterns of historical time-series data in the target database; the dynamic weight allocation model assigns corresponding weights to each prediction model based on the reliability of the prediction results output by each model; and the business-aware model understands the impact of business events on the capacity of the target database.

[0041] Optionally, the multi-dimensional features of the target database are input in parallel into multiple pre-built prediction models. Each prediction model infers the capacity trend of the target database from multiple dimensions based on its own design principles and outputs its own prediction results.

[0042] S203. Based on the prediction results output by each prediction model and the multi-dimensional characteristics of the target database, determine the target prediction results for the target database. The target prediction results include: the number of days the target database can be used and the confidence interval.

[0043] Optionally, the prediction results output by each prediction model are weighted and fused to obtain the initial prediction results for the target database. The initial prediction results are then adaptively adjusted by combining the contextual information provided by the business-related features in the multi-dimensional features of the target database to obtain the target prediction results for the target database.

[0044] The target prediction results include: number of days of available capacity and confidence interval. Number of days of available capacity refers to the number of days required for the remaining available capacity of the target database to be exhausted under the current trend; the confidence interval is a statistical interval used to characterize the range of uncertainty of the predicted value of the number of days of available capacity, providing a quantitative basis for risk assessment in operation and maintenance decisions.

[0045] In this embodiment, capacity metadata and performance index data of the target database are acquired, and historical time-series data of the target database is constructed according to the timestamps of the capacity metadata and performance index data. Multi-dimensional features of the target database, such as daily growth, long-term trend characteristics, periodic characteristics, business-related characteristics, and growth acceleration, are extracted from the historical time-series data. These multi-dimensional features are then input in parallel into multiple pre-built prediction models. Each prediction model performs capacity prediction for one dimension of the target database based on its built-in algorithm and outputs the corresponding prediction result. The prediction results output by each prediction model are weighted and fused to obtain the initial prediction result for the target database. The initial prediction result is then adaptively adjusted based on the business-related features of the target database's multi-dimensional features to obtain the target prediction result. The pre-built prediction models include a basic linear model, a periodic model, a dynamic weight allocation model, and a business-aware model, which can predict from multiple data dimensions, improving the accuracy of the prediction results. Furthermore, this application not only retains the basic linear model but also introduces a periodic model and a business-aware model to ensure that the prediction results for the target database can adapt to the non-linear growth and periodic fluctuations of the target database. Furthermore, this application adaptively adjusts the initial prediction results of the target database based on business association characteristics, thereby improving the accuracy of database capacity prediction. In this way, it can adapt to the non-linear growth and periodic fluctuations of the database, and adaptively adjust the prediction results based on the database's business data, thus improving the accuracy of database capacity prediction.

[0046] In one alternative implementation, see [link to implementation details]. Figure 3The specific operation of step S201 above can be as follows: S301. Obtain the connection information and data acquisition strategy of the target database.

[0047] Optionally, during the initialization phase of the database information prediction system, the connection information and data acquisition strategy of the target database are obtained. The connection information includes at least the target database's address, port, and service name. The data acquisition strategy defines the frequency, event window, and specific object list (such as a specific tablespace or specific database user) from which the database information prediction system acquires data from the target database.

[0048] S302. Establish a connection with the target database based on the connection information.

[0049] Optionally, after the database information prediction system obtains the connection information of the target database, it automatically calls standard database connection protocols (such as JDBC, ODBC, etc.) to establish a stable and secure network connection with one or more target databases based on the connection information of the target database.

[0050] S303. Collect the corresponding capacity metadata and performance index data of the target database from the data dictionary and dynamic performance view of the target database according to the data acquisition strategy, and compose historical time series data according to the timestamp of the capacity metadata and the timestamp of the performance index data.

[0051] The data dictionary is used to record the structural information of the target database, and the dynamic performance view is used to collect real-time statistics on the activity information of the target database.

[0052] Optionally, the database information prediction system schedules data acquisition tasks according to the pre-set data acquisition strategy of the target database, automatically queries the target database's data dictionary, and collects capacity metadata reflecting the target database's storage structure. Simultaneously, it queries the target database's dynamic performance view and collects performance indicator data reflecting the target database's operating status.

[0053] It is worth noting that while collecting data, the database information prediction system adds a precise timestamp to each data record and arranges the collected capacity metadata and performance index data in chronological order according to the timestamps to form historical time-series data of the target database.

[0054] S304. Preprocess the historical time series data and extract multi-dimensional features of the target database from the preprocessed historical time series data.

[0055] Optionally, to ensure the accuracy of intelligent analysis, the historical time-series data of the target database needs to be cleaned, transformed, and standardized to eliminate sudden interference from one-time data import.

[0056] Optionally, multi-dimensional features of the target database can be extracted from the processed historical time-series data to characterize the capacity changes of the target database from different dimensions.

[0057] In this way, historical time-series data is transformed into high-quality analytical features, providing direct and effective input data for subsequent predictive model analysis and reducing noise interference.

[0058] In one alternative implementation, see [link to implementation details]. Figure 4 The specific operation of step S304 above can be as follows: S401. Perform data cleaning and standardization on historical time series data to obtain structured time series data that meets the preset format.

[0059] Optionally, the collected raw historical time-series data can be preliminarily processed to eliminate noise and standardize data specifications.

[0060] Specifically, it checks and processes missing values ​​(e.g., by filling with interpolation from previous and next time points), corrects obvious formatting errors (e.g., character-to-numeric conversion), converts all capacity metadata to a uniform unit of measurement, and normalizes performance index data according to the definition, so that the raw historical time series data is transformed into clean, structured time series data that meets the preset format requirements.

[0061] S402. Perform outlier detection on the structured time series data to filter out abnormal data from the structured time series data.

[0062] Optionally, outlier detection is performed on the structured time series data, and the upper and lower quartiles and interquartile range of the structured time series data are calculated. Data that falls outside the reasonable range is identified as outlier data.

[0063] These abnormal data typically correspond to atypical capacity fluctuations caused by one-time data migration, batch testing, or temporary system failures, and this application does not specifically limit them.

[0064] It's important to note that identified outlier data is not simply deleted, but rather corrected using smoothing techniques. For example, it might be replaced with the mean, median, or moving average from adjacent time points to eliminate its interference with long-term trend analysis. This ensures the continuity and stationarity of the structured time series data, providing a more reliable data foundation for subsequent predictive models.

[0065] S403. Extract multi-dimensional features of the target database from the structured time-series data after filtering out abnormal data.

[0066] Optionally, feature extraction engineering is performed on the high-quality structured time-series data after cleaning, standardization, and anomaly handling to extract multi-dimensional features for capacity trend prediction.

[0067] In one optional implementation, the operation of step S402 may specifically include: determining the first data of the structured time series data at the first percentile position, and determining the second data of the structured time series data at the second percentile position.

[0068] Optionally, statistical analysis is performed on the structured time series data to obtain the first data (i.e., the lower quartile, denoted by Q1) at the first percentile position of the structured time series data, and the second data (i.e., the upper quartile, denoted by Q3) at the second percentile position of the structured time series data.

[0069] The first and second percentiles are values ​​preset by the user. Specifically, the first percentile can be 25% and the second percentile can be 75%. This application does not make any specific restrictions on these values.

[0070] For example, if the first percentile is 25% and the second percentile is 75%, the structured time-series data obtained after arranging them in chronological order can include 13 data points: Data 1, Data 2, Data 3, Data 4, Data 5, Data 6, Data 7, Data 8, Data 9, Data 10, Data 11, Data 12, and Data 13. The first data point at the first percentile (25%) is the fourth data point (Data 4), and the second data point at the second percentile (75%) is the tenth data point (Data 10). Each data point from Data 1 to Data 13 includes capacity metadata and performance metrics data at different time points.

[0071] It should be noted that structured time series data includes both capacity metadata and performance index data. Therefore, the first data taken at the first percentile position of structured time series data may be either capacity metadata or performance index data. Similarly, the second data taken at the second percentile position of structured time series data may be either capacity metadata or performance index data. This application does not make any specific limitations on this.

[0072] Furthermore, based on the first and second data, the interquartile range corresponding to the structured time series data is determined.

[0073] Optionally, the interquartile range (IQR) corresponding to the structured time series data can be calculated using the following formula (1): IQR = Q3 - Q1 (1) Wherein, IQR is used to represent the interquartile range, Q1 is used to represent the first data in the structured time series data at the first percentile position, and Q3 is used to represent the second data in the structured time series data at the second percentile position.

[0074] For example, if the first data is data 4 and the second data is data 10, then the interquartile range (IQR) corresponding to the structured time series data is data 10 - data 4.

[0075] Furthermore, based on the first data and the interquartile range, the lower bound of outlier data corresponding to the structured time series data is determined.

[0076] Optionally, the lower bound of outlier data corresponding to structured time series data can be calculated using the following formula (2), which is as follows: Lower=Q1-k IQR (2) Where Lower represents the lower bound of outlier data, Q1 represents the first data point in the structured time series data at the first percentile, IQR represents the interquartile range, and k represents the outlier coefficient.

[0077] It is worth noting that the anomaly coefficient is a pre-set threshold, and the anomaly coefficient can be 1.5. This application does not make a specific limitation on this.

[0078] For example, if the anomaly coefficient k is 1.5, the interquartile range (IQR) corresponding to the structured time series data is data10 - data4, and the first data in the structured time series data at the first percentile position is data4, then the lower bound of the anomaly data corresponding to the structured time series data is data4 - 1.5 × (data10 - data4).

[0079] It should be noted that the data contained in the structured time series data are sorted only according to their respective timestamps, regardless of the data size. The first data is not necessarily always greater than the second data, nor is the first data always less than the second data. This application does not make any specific limitations in this regard.

[0080] Furthermore, based on the second data and the interquartile range, the upper bound of the outlier data corresponding to the structured time series data is determined.

[0081] Optionally, the upper bound of outlier data corresponding to structured time series data can be calculated using the following formula (3), which is as follows: Upper = Q3 + k IQR (3) Where Upper represents the upper bound of outlier data, Q3 represents the second data point in the structured time series data at the second percentile, IQR represents the interquartile range, and k represents the outlier coefficient.

[0082] For example, if the anomaly coefficient k is 1.5, the interquartile range (IQR) corresponding to the structured time series data is data 10 - data 4, and the second data at the second percentile position of the structured time series data is data 10, then the upper bound of the anomaly data corresponding to the structured time series data is data 10 + 1.5 × (data 10 - data 4).

[0083] Furthermore, based on the upper and lower bounds of outlier data, outlier data is filtered out from structured time-series data.

[0084] Optionally, each data point in the structured time series data is compared with the lower bound and the upper bound of outlier data. Any data point that is less than the lower bound or greater than the upper bound of outlier data is identified as outlier data.

[0085] Furthermore, the database information prediction system records the location and value of these abnormal data points for subsequent smoothing or specialized analysis.

[0086] For example, if the upper bound of the abnormal data corresponding to the structured time series data is Upper = data10 + 1.5 × (data10 - data4), and the lower bound of the abnormal data is Lower = data4 - 1.5 × (data10 - data4), then all data in the 13 data from data1 to data13 contained in the structured time series data that are less than the lower bound of the abnormal data and greater than the upper bound of the abnormal data will be removed as abnormal data.

[0087] In this embodiment of the application, "anomalies" are defined by the distribution characteristics of the data itself, which can effectively identify and filter out capacity data points that deviate from the normal growth pattern due to sudden business operations (such as large-scale data import) or temporary system failures, thereby ensuring the accuracy of subsequent trend analysis and feature extraction.

[0088] In one optional implementation, the aforementioned multiple prediction models may include: a basic linear model, a periodic model, and a dynamic weight allocation model, see [link to relevant documentation]. Figure 5 The specific operation of step S202 above can be as follows: S501. Input the daily growth rate, growth acceleration and long-term trend features of the data in the multi-dimensional features into the basic linear model to obtain the linear prediction results of the target database. The linear prediction results are used to characterize the short-term data change trend of the target database.

[0089] Optionally, the daily growth rate, growth acceleration, and long-term trend characteristics of the target database's multi-dimensional features are input into a base linear model (such as a Baseline Trend model). The base linear model uses algorithms such as linear regression to fit the overall slope and acceleration of the target database's capacity change over time, outputting a linear prediction result for the target database. This linear prediction result primarily characterizes the short-term stable trend of the target database's capacity, ignoring short-term fluctuations, providing a baseline for the target database's prediction results.

[0090] S502. Input the periodic features from the multi-dimensional features into the periodic model to obtain the periodic prediction results of the target database. The periodic prediction results are used to characterize the periodic data change trend of the target database.

[0091] Optionally, the periodic features (especially seasonal components on a weekly or monthly basis) from the multidimensional features of the target database can be input into a periodic model (such as a model based on STL time series decomposition). The periodic model is used to identify and extrapolate recurring periodic patterns in historical time series data to output periodic prediction results for the target database. These periodic prediction results characterize the regular fluctuation trend of the target database capacity due to business rhythms, such as the periodic changes between weekend troughs and weekday peaks.

[0092] S503. Input the linear prediction results and the periodic prediction results into the dynamic weight allocation model to obtain the target weight sequence, which includes the weight values ​​of the basic linear model and the weight values ​​of the periodic model.

[0093] Optionally, the linear and periodic prediction results can be input into a dynamic weighting model. The dynamic weighting model analyzes the prediction performance of the basic linear and periodic models on recent historical time series data. Evaluation metrics may include prediction error (such as mean squared error), goodness of fit, or the model's own confidence score.

[0094] Furthermore, based on the performance evaluation results, the dynamic weight allocation model dynamically calculates and assigns the relative importance of the basic linear model and the periodic model within the current prediction period, generating a target weight sequence. This weight sequence includes the weight values ​​of the basic linear model and the weight values ​​of the periodic model.

[0095] It should be noted that models with better performance and higher confidence levels will be assigned higher weights.

[0096] In one optional implementation, the aforementioned multiple prediction models may further include: a business-aware model, see [link to relevant documentation]. Figure 6 The specific operation of step S203 above can be as follows: S601. Based on the target weight series, the linear prediction results and the periodic prediction results are weighted and fused to obtain the initial prediction results of the target database.

[0097] Optionally, the linear prediction results and the periodic prediction results are weighted and fused based on the target weight sequence output by the dynamic weight allocation model.

[0098] Optionally, the initial prediction result of the target database can be determined by the following formula (4), which is as follows: P=w1 L+w2 S(4) Where P represents the initial prediction result of the target database, w1 represents the weight value of the basic linear model, L represents the linear prediction result, w2 represents the weight value of the periodic model, and S represents the periodic prediction result.

[0099] Therefore, the initial forecast results mathematically combine short-term linear trends with cyclical fluctuations, making it a more robust baseline forecast.

[0100] S602. Input the business-related features from the multi-dimensional features and the initial prediction results of the target database into the business-aware model to obtain the target prediction results of the target database.

[0101] Optionally, the business-related features (such as correlation indicators with business volume and promotional activities) from the multi-dimensional features of the target database, along with the initial prediction results of the target database, are jointly input into the business-aware model. The business-aware model performs semantic parsing on the business-related features to identify known business plans (such as large-scale marketing campaigns and financial report data generation), holiday arrangements, and historical business peak patterns. Based on predefined business logic rules or learned adjustment patterns, the model adaptively adjusts the initial prediction results and outputs the target prediction results after business semantic calibration.

[0102] It is worth noting that the target forecast results not only include a forecast of the number of days of available capacity that is closer to business reality, but also usually include a confidence interval that reflects the uncertainty of this forecast. This confidence interval may be appropriately adjusted based on the confidence interval of the initial forecast due to the complexity of business events.

[0103] In one alternative implementation, see [link to implementation details]. Figure 7 The specific operation of step S602 above can be as follows: The business awareness model acquires external business calendar information related to the target database and parses this information to determine key business time nodes such as peak business periods, statutory holidays, and known pending business events within the future forecast period. Peak business periods refer to expected peak load periods caused by business patterns (such as month-end settlements, e-commerce promotions, and financial report releases); statutory holidays refer to national statutory holidays such as the Mid-Autumn Festival and National Day; and known pending business events include planned business operations that may significantly impact data volume, such as large-scale data migration and historical data archiving.

[0104] S701. Determine the sequence of data change influencing factors for the target database in the future period. The sequence of data change influencing factors includes: influencing factors during peak business periods, influencing factors during statutory holidays, and influencing factors for pending business events.

[0105] Optionally, based on the identified business time points, the business perception model calculates quantified impact factors for each, forming an impact factor sequence. The calculation of each impact factor is based on historical time-series data analysis.

[0106] Specifically, the impact factor during peak business periods is usually a negative adjustment coefficient (e.g., -0.1 to -0.3), indicating that the daily capacity consumption of the target database during this period may increase by % to 30% compared to normal days; the impact factor during statutory holidays is usually a positive adjustment coefficient (e.g., +0.05 to +0.2), indicating that the daily capacity consumption of the target database during this period may decrease by 5% to 20% compared to normal days, but it may also be negative (e.g., holiday promotional activities); the impact factor for pending business events is calculated based on the type of business event and the estimated data volume, using a specific adjustment value (e.g., a one-time increase of 50GB) or adjustment coefficient.

[0107] S702. Based on the business perception model, adjust the initial prediction results of the target database according to the data change influencing factor sequence to obtain the target prediction results of the target database.

[0108] Optionally, the business-aware model adaptively adjusts the initial prediction results of the target database based on the influencing factor sequence to obtain the target prediction results for the target database.

[0109] In this way, the target prediction results output by the database information processing system are more in line with the actual business operation rhythm, significantly improving the prediction accuracy at key business time points and providing intelligent suggestions with business interpretability for capacity planning.

[0110] In one alternative implementation, see [link to implementation details]. Figure 8 The aforementioned database information prediction method also includes: S801. Determine the health score of the target database based on the target prediction results of the target database and the weight parameters of the preset database type.

[0111] Optionally, the health score of the target database is calculated based on the target prediction results of the target database and the weight parameters corresponding to the pre-configured database type.

[0112] Specifically, based on the number of days of available capacity in the target forecast results, from Figure 12 The provided capacity utilization rating mapping table is used to find the score corresponding to the number of days of available capacity (e.g., if the target database has 200 days of available capacity, the corresponding score is 70). Based on the historical capacity growth data contained in the historical time-series data of the target database, the fluctuation of the target database's growth rate is determined to assess the stability of the target database, and the calculated growth rate is used to determine the target database's stability. Figure 13 Find the frequency score corresponding to the growth rate in the provided growth rate trend score mapping table.

[0113] Optionally, the database information prediction system adjusts the corresponding weight parameters based on the target database's database type (such as an OLIP online transaction database, an OLAP analytical database, or a core business database). These weight parameters define the weighting of capacity utilization and capacity growth rate trends in the health score (e.g., OLAP databases have a greater emphasis on the capacity dimension).

[0114] Optionally, the health score of the target database can be determined based on the weight parameters of the target database, the available days of capacity score, and the growth rate trend score of the target database.

[0115] S802. Based on the health score results and target prediction results of the target database, perform graded early warning or automatic expansion on the target database respectively.

[0116] Optionally, differentiated operation and maintenance decisions can be automatically executed based on the health score results and target prediction results of the target database.

[0117] Specifically, the database information prediction system determines the business weight of the target database based on its importance level in the business system, determines the capacity availability level based on the number of available days in the target prediction results of the target database, and determines the stability of the growth trend of the target database based on its historical time series data, so as to obtain the health score result of the target database.

[0118] Furthermore, queries can be performed based on health score results from the target database. Figure 14 The provided decision matrix mapping table determines the corresponding decision output and automatically triggers the corresponding operation.

[0119] In one alternative implementation, see [link to implementation details]. Figure 9 The specific operation of step S801 above can be as follows: S901. Based on the number of days of available capacity in the target prediction results, determine the capacity utilization score of the target database.

[0120] Optionally, query based on the number of days of available capacity in the target prediction results of the target database. Figure 12 The provided capacity utilization score mapping table determines the target database's score in the capacity utilization dimension, i.e., determines the target database's capacity utilization score.

[0121] Specifically, according to Figure 12 According to the publicly available information, when the predicted number of days the target database's capacity is available is greater than 180 days, the target database's capacity utilization score is 70 points, and the risk level of the target database in the capacity utilization dimension is safe; when the predicted number of days the target database's capacity is available is between 30 and 90 days, the target database's capacity utilization score is 50 points, and the risk level of the target database in the capacity utilization dimension is good; when the predicted number of days the target database's capacity is available is between 30 and 40 days, the target database's capacity utilization score is 30 points, and the risk level of the target database is... If the risk level in the capacity utilization dimension is "Attention Level," then it is necessary to plan for expansion of the target database. If the predicted number of days the target database has available capacity is between 2 and 30 days, the target database's capacity utilization score is 20 points, and the risk level in the capacity utilization dimension is "Dangerous Level," then it is necessary to immediately expand the target database. If the predicted number of days the target database has available capacity is less than 2 days, the target database's capacity utilization score is 0 points, and the risk level in the capacity utilization dimension is "Critical Level," then it is necessary to urgently expand the target database.

[0122] In this way, the risk of the predicted number of days of available capacity for the target database can be quantified, and the predicted number of days of available capacity can be converted into a comparable score.

[0123] S902. Based on multi-dimensional features, the historical capacity growth data of the target database is obtained by fitting, and the growth rate trend score of the target database is determined based on the historical capacity growth data.

[0124] Optionally, based on the multi-dimensional features extracted from the historical time-series data of the target database, the fluctuation characteristics of the historical capacity growth data of the target database can be fitted through statistical analysis (such as calculating the standard deviation of the growth rate, the coefficient of variation, or the stationarity of the fitted trend).

[0125] Furthermore, the stationarity level of the target database is assessed based on its fluctuation characteristics, and based on... Figure 13 The provided growth rate trend score mapping table determines the growth rate trend score results for the target database.

[0126] Specifically, according to Figure 13 According to the publicly available information, when the growth rate represented by the historical capacity growth data of the target database is ±%, the growth rate trend score of the target database is 30 points, and no analysis is needed; when the growth rate represented by the historical capacity growth data of the target database is ±30%, the growth rate trend score is 20 points, and attention is needed; when the growth rate represented by the historical capacity growth data of the target database is ±50%, the growth rate trend score is [missing information], and analysis is needed; when the growth rate represented by the historical capacity growth data of the target database is ±0%, the growth rate trend score is 0 points, indicating excessive business fluctuations in the target database, requiring immediate action.

[0127] S903. Based on the database type of the target database, determine the corresponding weight parameters for the target database. The weight parameters are used to indicate the importance of the capacity utilization rate and growth rate trend of the target database in the health score results.

[0128] Optionally, based on the database type of the target database, a pre-configured set of weights (see [link to relevant documentation]) can be selected. Figure 10 The corresponding weight parameters are retrieved and determined from the database.

[0129] It should be noted that this weighting parameter is a vector or dictionary that explicitly indicates the relative importance of the capacity utilization rate and growth rate trend in the final health score calculation. For example, for OLAP databases, the weight of the capacity dimension (e.g., 0.5) will be higher than that for OLTP databases.

[0130] S904. Based on the capacity utilization rate score, the growth rate trend score, and the weighting parameters, determine the health score of the target database.

[0131] Optionally, the health score results of the target database can be determined by the following formula (5), which is as follows: H = X w3 + Y × w4 + W (5) Wherein, H represents the health score of the target database, X represents the capacity utilization score, Y represents the growth rate trend score, w3 and w4 are the weight parameters of the target database, w3 represents the weight coefficient of capacity utilization, w4 represents the weight coefficient of growth rate trend, and W represents the preset adjustment factor.

[0132] Optionally, the adjustment factor can be fine-tuned based on additional conditions such as whether there are seasonal peaks or whether automatic expansion capabilities are available.

[0133] In addition, the health score (any value between 0 and 0) of the target database is output along with the risk level, providing an intuitive and comprehensive evaluation basis for subsequent operation and maintenance decisions.

[0134] In one alternative implementation, see [link to implementation details]. Figure 11 The specific operation of step S802 above can be as follows: S1101. Obtain the preset business weights of the target database.

[0135] Optionally, the preset business weights of the target database at the business level can be obtained. These preset business weights are usually configured during the initialization of the database information prediction system or during business relationship mapping, and are used to identify the importance level of the business carried by the target database.

[0136] S1102. Based on the number of days of available capacity in the target database's target prediction results, determine the corresponding level of available capacity days for the target database.

[0137] Optionally, based on the number of days of available capacity in the target prediction results of the target database, in Figure 12 The risk level corresponding to the number of days of available capacity is found in the provided mapping table, and the found risk level is used as the level of available capacity days in the target database.

[0138] S1103. Based on the growth rate trend score in the health score of the target database, determine the growth trend stability level of the target database.

[0139] Optionally, based on the growth rate trend score results in the health score results of the target database, from Figure 13 The system finds the corresponding intelligent suggestions in the provided mapping table for the growth rate trend score, and uses the found intelligent suggestions as the growth trend stability level (such as stable, irregular, etc.) for the target database.

[0140] S1104. Based on the preset business weights, capacity availability days levels, and growth trend stability levels of the target database, query the preset decision matrix to obtain the target decision corresponding to the target database.

[0141] Optionally, the preset decision matrix is ​​a matrix pre-established and maintained by the database information prediction system. This decision matrix takes the business weight, capacity availability days level, and growth rate trend stability level of the target database as joint input conditions, and can output the corresponding target decision.

[0142] S1105. Perform corresponding operations based on the target decision corresponding to the target database, wherein the target decision includes at least one of the following: display only the prediction results, proactively notify and warn, or trigger automatic expansion.

[0143] Optionally, based on the target decision obtained from the query, the system automatically executes the corresponding operations to achieve closed-loop operation and maintenance. The target decisions include: displaying only the prediction results, proactively issuing warnings, and triggering automatic capacity expansion.

[0144] Specifically, "displaying only the prediction results" means updating the target prediction results and health score results of the target database in the visual monitoring interface of the database information prediction system without external notification; "proactive notification and early warning" means generating alarm information containing target prediction results, health score results, and risk level analysis, and proactively pushing it to authorized operation and maintenance personnel through integration channels; "triggered automatic expansion" means that the database information prediction system automatically calls predefined expansion scripts or cloud platform APIs to perform online storage expansion operations (such as expanding tablespaces or adding data files), and verifies the expansion results after the expansion operation is completed to update the database's capacity metadata.

[0145] Figure 14 A mapping table of a decision matrix provided in this application, see [link to table]. Figure 14 The specific operation of step S14 above can be as follows: S1: If the preset business weight of the target database is low, the target decision of the target database is to only display the prediction results; conversely, if the preset business weight of the target database is high, the target decision of the target database is to actively notify and warn. S2: If the number of days of available capacity in the target prediction results of the target database is less than 30 days, the target decision of the target database is to only display the prediction results. If the number of days of available capacity in the target prediction results of the target database is between 2 and 30 days, the target decision of the target database is to actively notify and warn. If the number of days of available capacity in the target prediction results of the target database is less than 2 days, the target decision of the target database is to trigger automatic expansion. S3: If the growth rate trend stability level of the target database is stable, the target decision of the target database is to only display the prediction results. If the growth rate trend stability level of the target database is irregular, the target decision of the target database is to actively notify and warn.

[0146] The following describes the computer equipment and computer-readable storage medium used to execute the database information prediction method provided in this application. The specific implementation process and technical effects are described above and will not be repeated below.

[0147] Figure 15 This is a schematic diagram of the structure of a computer device provided in this application. See also: Figure 15 The computer device includes: a memory 1501 and a processor 1502. The memory 1501 stores a computer program that can run on the processor 1502. When the processor 1502 executes the computer program, it implements the steps in any of the above method embodiments.

[0148] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps in the various method embodiments described above.

[0149] Optionally, this application also provides a program product, such as a computer-readable storage medium, including a program that, when executed by a processor, performs any of the above-described database information prediction method embodiments.

[0150] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute partial steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0151] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0152] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A database information prediction method, characterized in that, The method includes: Based on the capacity metadata and performance metrics of the target database, the multi-dimensional characteristics of the target database are determined. These multi-dimensional characteristics include: daily data growth, long-term trend characteristics, periodic characteristics, business association characteristics, and growth acceleration. Based on the multi-dimensional features of the target database and multiple pre-built prediction models, the prediction results output by each prediction model are obtained. Based on the prediction results output by each prediction model and the multi-dimensional features of the target database, the target prediction result of the target database is determined. The target prediction result includes: the number of days of available capacity of the target database and the confidence interval.

2. The database information prediction method according to claim 1, characterized in that, The step of determining the multi-dimensional characteristics of the target database based on its capacity metadata and performance metrics includes: Obtain the connection information and data acquisition strategy of the target database; Establish a connection with the target database based on the connection information; According to the data acquisition strategy, the capacity metadata and performance index data corresponding to the target database are collected from the data dictionary and dynamic performance view of the target database, and historical time-series data are composed according to the timestamp of the capacity metadata and the timestamp of the performance index data. The historical time-series data is preprocessed, and multi-dimensional features of the target database are extracted from the preprocessed historical time-series data.

3. The database information prediction method according to claim 2, characterized in that, The preprocessing of the historical time-series data and the extraction of multi-dimensional features of the target database from the preprocessed historical time-series data include: The historical time series data is cleaned and standardized to obtain structured time series data that meets the preset format; Anomaly detection is performed on the structured time series data to filter out abnormal data from the structured time series data; Multidimensional features of the target database are extracted from the structured time-series data after filtering out abnormal data.

4. The database information prediction method according to claim 3, characterized in that, The outlier detection process for the structured time series data, in order to filter out abnormal data from the structured time series data, includes: Determine the first data at the first percentile position of the structured time series data, and determine the second data at the second percentile position of the structured time series data; Based on the first data and the second data, determine the interquartile range corresponding to the structured time series data; Based on the first data and the interquartile range, determine the lower bound of the abnormal data corresponding to the structured time series data; Based on the second data and the interquartile range, determine the upper bound of the abnormal data corresponding to the structured time series data; Abnormal data is filtered out from the structured time-series data based on the upper and lower bounds of the abnormal data.

5. The database information prediction method according to claim 1, characterized in that, The multiple prediction models include: a basic linear model, a periodic model, and a dynamic weight allocation model; The step of obtaining the prediction results output by each prediction model based on the multi-dimensional features of the target database and multiple pre-built prediction models includes: The daily growth rate, growth acceleration, and long-term trend characteristics of the data in the multi-dimensional features are input into the basic linear model to obtain the linear prediction results of the target database. The linear prediction results are used to characterize the short-term data change trend of the target database. The periodic features in the multi-dimensional features are input into the periodic model to obtain the periodic prediction results of the target database. The periodic prediction results are used to characterize the periodic data change trend of the target database. The linear prediction results and the periodic prediction results are input into the dynamic weight allocation model to obtain a target weight sequence, which includes the weight values ​​of the basic linear model and the weight values ​​of the periodic model.

6. The database information prediction method according to claim 5, characterized in that, The multiple prediction models also include: a business-aware model; The step of determining the target prediction result of the target database based on the prediction results output by each prediction model and the multi-dimensional features of the target database includes: The linear prediction results and the periodic prediction results are weighted and fused based on the target weight sequence to obtain the initial prediction results for the target database. The business-related features from the multi-dimensional features and the initial prediction results of the target database are input into the business-aware model to obtain the target prediction results of the target database.

7. The database information prediction method according to claim 6, characterized in that, The step of inputting the business-related features from the multi-dimensional features and the initial prediction results of the target database into the business-aware model to obtain the target prediction results of the target database includes: Determine the sequence of data change influencing factors for the target database in the future period. The sequence of data change influencing factors includes: influencing factors during peak business periods, influencing factors during statutory holidays, and influencing factors for pending business events. Based on the business perception model, the initial prediction results of the target database are adjusted according to the data change influencing factor sequence to obtain the target prediction results of the target database.

8. The database information prediction method according to claim 1, characterized in that, The method further includes: Based on the target prediction results of the target database and the weight parameters of the preset database type, the health score result of the target database is determined; Based on the health score and target prediction results of the target database, tiered early warning or automatic expansion is performed on the target database respectively.

9. The database information prediction method according to claim 8, characterized in that, The step of determining the health score result of the target database based on the target prediction result of the target database and the weight parameters of the preset database type includes: Based on the number of days of available capacity in the target prediction results, the capacity utilization score of the target database is determined; Based on the multi-dimensional features, historical capacity growth data of the target database is obtained by fitting, and the growth rate trend score of the target database is determined based on the historical capacity growth data. Based on the database type of the target database, a weight parameter corresponding to the target database is determined. The weight parameter is used to indicate the importance of the capacity utilization rate and growth rate trend of the target database in the health score result. The health score of the target database is determined based on the capacity utilization score, the growth rate trend score, and the weighting parameters.

10. A computer device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 9.

Images