A unified data management system and method based on machine learning

Abstract

The present invention discloses a unified data management system and method based on machine learning, and belongs to the field of data analysis technology. The present invention manually reviews resource data that has passed through an automated classification system in a data cloud platform, collects and analyzes the number of errors for which the review results of various resource data are errors, and obtains several time intervals; establishes an error number prediction model for various resource data collected at different time intervals, and tests to obtain the optimal prediction model for the number of errors for various resource data; calculates several combinations of the number of errors for various resource data based on a threshold of the total time spent by staff members reviewing the error data; and calculates the update interval duration based on several combinations of the number of errors for various resource data. The automated classification system is updated in a timely manner based on the update interval duration to reduce the occurrence of errors.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a unified data management system and method based on machine learning. Background Art

[0002] Today, big data cloud platform technology is rapidly developing, replacing previous manual record-keeping methods. Unified cloud management platforms are already in use by most companies. Unified cloud management platforms are integrated cloud management solutions designed to centrally manage and monitor diverse cloud resources through a unified interface and management tools. By using unified cloud management platforms, enterprises can more efficiently manage and monitor cloud resources, optimize resource usage, and ensure the security and compliance of their cloud environments.

[0003] In the managed data cloud platform, there is an automated classification system that automatically classifies and summarizes various resource data. When users upload resource data to the cloud platform, the resource data is automatically classified by the automated classification system and placed in the type to which it belongs. However, since resource data will change over time, the automated classification system is not updated in a timely manner, which will cause the automated classification system to make errors in classification judgments and place resource data belonging to this category in other categories. Staff need to re-examine and place the resource data in the correct location. This operation affects the time of manual review and has an impact on the work efficiency of staff. Therefore, a unified managed data management system and method based on machine learning is needed to obtain the update interval duration, and update the automated classification system in a timely manner according to the update interval duration to reduce errors. Summary of the Invention

[0004] The purpose of the present invention is to provide a unified data management system and method based on machine learning to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a unified data management system and method based on machine learning, the method comprising:

[0006] Step S100: In the managed data cloud platform, manually review the resource data that has been through the automated classification system, collect and analyze the number of errors in the resource data reviewed as errors, and obtain a number of time intervals;

[0007] Step S200: establishing an optimal prediction model for the number of errors in various resource data based on a number of time intervals;

[0008] Step S300: obtaining a plurality of combinations of error quantities of various resource data and a set of time durations for the plurality of combinations of error quantities according to a threshold of the total time spent by staff members reviewing error data;

[0009] Step S400: Calculate the optimal level of the time spent based on a set of time spent for several combinations of error quantities, and update and analyze the automated classification system based on the range of the optimal level.

[0010] Furthermore, step S100 includes:

[0011] Step S101: manually review the i-th resource data, collect the number of errors that are reviewed as errors, and plot a curve chart of the number of errors in the i-th resource data;

[0012] Step S102: Calculate the curve graph to obtain the change period of the i-th resource data, aggregate and calculate the change periods of various resource data in the managed data cloud platform, and obtain the average A and median B of the resource data change periods;

[0013] Step S103: Calculate the time interval for auditing resource data based on the mean and median, and calculate the time interval formula: G n =A+n×B, where G n It represents the nth time interval, n represents the weight of the time interval, and n is randomly selected within the specified range. The random value is input into the time interval formula to obtain several time intervals.

[0014] In the above steps, the number of errors with review results as errors is collected based on the number of new errors in various resource data at different time points. This is because after each review, the erroneous resource data will be put back to its correct category.

[0015] Since artificially set time intervals are prone to data redundancy or decreased prediction accuracy, a more appropriate time interval can be found by summarizing the number of errors in resource data using a curve chart.

[0016] The horizontal axis of the curve graph is the time point of collection, and the vertical axis is the number of errors in the resource data. According to the analysis of the curve graph, starting from the origin of the curve graph, the time period when the slope value changes twice is calculated to obtain the change period of the resource data;

[0017] Take the mean and median of the change cycle, substitute them into the time interval formula, and randomly select the weight of the time interval within the range to obtain several random time intervals. This can make the time interval richer and obtain more appropriate time intervals. If the weight of the time interval is not randomly selected, the time interval will be too single, which may easily cause errors in subsequent analysis.

[0018] Furthermore, step S200 includes:

[0019] Step S201: G n As the time interval, the number of errors in the i-th resource data is collected, and a G n For the time interval, the error number prediction model of the i-th resource data is tested, and the prediction model is obtained by G n For the time interval, the error value of the error quantity prediction model of the i-th resource data will be expressed as G n For the time interval, the error values ​​of the error quantity prediction models of various resource data are added together to obtain the G n is the total error value of the time interval;

[0020] Step S202: Retrieve several time intervals, summarize the total error values ​​of each time interval, sort them from small to large, take the time interval corresponding to the minimum value as the optimal time interval, and use the error quantity prediction model of various resource data corresponding to the optimal time interval as the optimal error quantity prediction model of various resource data.

[0021] In the above steps, a prediction model is established by using a linear regression model based on the time interval and the number of errors collected;

[0022] The number of errors in resource data is collected according to several time intervals, and a prediction model for the number of errors in resource data is established. The error values ​​of the prediction models obtained at different time intervals are judged to obtain the optimal prediction model. If the number of errors in resource data is not collected at multiple time intervals, the accuracy of the prediction model will be too low, which will affect subsequent analysis.

[0023] Furthermore, step S300 includes:

[0024] Step S301: Collect the time spent by staff on manual review of resource data, and calculate the average time spent by staff on manual review of resource data as T p , according to the average time, set a threshold T of the total time spent by staff members reviewing erroneous data;

[0025] Step S302: Collect the average time spent by staff on reviewing the error data of the i-th resource data, and calculate the total time spent on reviewing the number of errors in various resource data. According to the total time spent on review formula: T'=T1×S1+T2×S2+...+T i ×S i , where T' represents the total time spent on the review, T1, T2, ..., T i They are the average time spent on error data review for the 1st, 2nd, ..., i-th resource data, S1, S2, ..., S i They are represented as the number of errors in the 1st, 2nd, ..., i-th resource data respectively;

[0026] Step S303: The jth error quantity combination that satisfies the total time spent on reviewing is lower than the total time spent on reviewing error data is S 1j 、S 2j ,...,S ij , where S 1j 、S 2j ,...,S ij They are respectively represented as the number of errors of the 1st, 2nd, ..., i-th resource data in the j-th error number combination;

[0027] Step S304: S ij Input into the optimal prediction model of the number of errors of the i-th resource data, and the number of errors of the i-th resource data reaches S ij The time taken is to summarize the time taken for the number of errors of various resource data to reach the number of errors of various resource data in the jth error number combination to obtain the time set taken for the jth error number combination.

[0028] Since the time staff spend on reviewing classified resource data varies from day to day, the average time spent by staff is calculated to obtain a threshold for the total time staff spend reviewing erroneous data. The threshold can be adjusted based on the weight of the time staff spend reviewing erroneous data to obtain a more appropriate threshold. If m is larger, T is also larger, then staff can accept more total time spent reviewing erroneous data. If m is smaller, T is also smaller, then staff can accept less total time spent reviewing erroneous data. m is assigned a value according to different situations to obtain an appropriate threshold for the total time spent reviewing erroneous data.

[0029] The calculation is based on the threshold of the total time spent by staff members reviewing erroneous data and the average time spent reviewing erroneous data for each type of resource data. Since there are several types of resource data and the average time spent reviewing each type of resource data is different, there will be several possible combinations of erroneous quantities for the resource data that meet the error quantity formula. If each combination of erroneous quantities is not calculated, the combination will be simplified, resulting in errors in subsequent calculations and inaccurate results, which will affect subsequent analysis.

[0030] According to each combination of error numbers, the number of errors in each resource data is input into the corresponding optimal prediction model to obtain the predicted duration, and the average duration of each combination of error numbers is calculated. The duration corresponding to each combination of error numbers is different, so the duration corresponding to each possibility must be calculated to avoid inaccurate results due to too little data.

[0031] Furthermore, step S400 includes:

[0032] Step S401: Retrieve a set of time spent for various combinations of error numbers, input the set of time spent into a K-means clustering algorithm, calculate the sum of squared errors of the K-means clustering, observe the change of the sum of squared errors using the elbow rule, and determine that the time spent has r levels;

[0033] Step S402: comparing the number of times spent in each level, and taking the level with the largest number of times spent as the optimal level;

[0034] Step S403: When the time between the last update time and the present time of the automated classification system reaches the minimum value of the optimal level, an update reminder is issued to the automated classification system. When the time between the last update time and the present time of the automated classification system reaches the maximum value of the optimal level, the automated classification system is shut down and an alarm is issued.

[0035] The update reminder duration is classified and determined based on the obtained set of spent durations in order to avoid manual decisions, which may lead to subjective determination of the reminder duration. Most of the spent durations obtained according to the optimal level are within the optimal level range, so determining the update reminder duration based on the optimal level range will reduce the occurrence of errors.

[0036] In order to better implement the above method, a unified data management system based on machine learning is proposed. The system includes a time interval module, a prediction model module, a quantity combination module, and an update duration module.

[0037] Time interval module: In the Naguan Data Cloud Platform, resource data that has been automatically classified by the system is manually reviewed. The number of errors identified in the review results for various resource data is collected and analyzed to obtain several time intervals.

[0038] Prediction model module: establishes the optimal prediction model for the number of errors in various resource data based on several time intervals;

[0039] Quantity combination module: Based on the threshold of the total time spent by staff members reviewing error data, several error quantity combinations of various resource data and the time spent on several error quantity combinations are obtained;

[0040] Update duration module: Calculates the optimal level of the time spent based on a set of time spent for several combinations of error quantities, and performs update analysis on the automated classification system based on the range of the optimal level.

[0041] Furthermore, the time interval module includes a collection time interval unit and a calculation time interval unit;

[0042] Collection time interval unit: manually review various resource data, collect the number of errors that are reviewed as errors, and draw a curve chart of the number of errors in various resource data;

[0043] Calculation time interval unit: Analyze the curve graph to obtain the change cycle of various resource data curve graphs, perform weighted calculation, and obtain several time intervals.

[0044] Furthermore, the prediction model module includes a prediction model establishment unit and an optimal prediction model analysis unit;

[0045] Establishing a prediction model unit: collecting the number of errors in various resource data at different time intervals, and establishing a prediction model for the number of errors in various resource data collected at different time intervals;

[0046] Analyze the optimal prediction model unit: train the prediction model to obtain the total error value of the error quantity prediction model of various resource data collected at different time intervals, and take the prediction model with the smallest total error value as the optimal prediction model.

[0047] Furthermore, the quantity combination module includes an error time analysis unit and an error quantity combination unit;

[0048] Error time analysis unit: The time spent by staff on manual review of resource data is collected to obtain the average time. The average time is weighted and calculated to obtain the threshold of the total time spent by staff on reviewing error data.

[0049] Error quantity combination unit: Calculate based on the threshold of the total time spent by staff on reviewing error data and the average time spent on reviewing error data of various resource data to obtain several combinations of error quantities of various resource data. Summarize the time it takes for the error quantity of various resource data to reach the error quantity of various resource data in various error quantity combinations to obtain a set of time spent on various error quantity combinations.

[0050] Furthermore, the update duration module includes an update duration analysis unit and an update reminder unit;

[0051] Analyze and update the duration unit: obtain various levels of the duration according to the duration of various error number combinations, and take the duration level with the largest number of errors as the optimal level;

[0052] Update reminder unit: When the time between the last update time and the present time of the automated classification system reaches the minimum value of the optimal level, an update reminder will be issued to the automated classification system. When the time between the last update time and the present time of the automated classification system reaches the maximum value of the optimal level, the automated classification system will be shut down and an alarm will be issued.

[0053] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: the present invention manually reviews the resource data that has passed the automated classification system in the managed data cloud platform, collects and analyzes the number of errors in the various resource data that are reviewed as errors, and obtains several time intervals. The above method effectively avoids the irregularity of the resource data collected due to artificially set time intervals, and multiple time intervals can support the subsequent acquisition of a more accurate prediction model; establishes a prediction model for the number of errors in various resource data collected at different time intervals, and tests to obtain the optimal prediction model for the number of errors in various resource data; the above method obtains the optimal prediction model by testing each prediction model, avoiding only one prediction model, resulting in incorrect data and affecting subsequent analysis; according to the threshold of the total time spent by the staff on reviewing the erroneous data, calculates several combinations of the number of errors in various resource data, and the above method avoids the simplification of the number of errors in various resource data, thereby improving the accuracy of the subsequent update time; according to several combinations of the number of errors in various resource data, the update interval is calculated, and the automated classification system is updated in time according to the update interval to reduce the occurrence of errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0055] Figure 1This is a flow chart of a unified data management method based on machine learning in the present invention;

[0056] Figure 2 This is a structural diagram of a unified data management system based on machine learning in the present invention;

[0057] Figure 3 It is a schematic diagram of an embodiment of a unified data management method based on machine learning of the present invention. DETAILED DESCRIPTION

[0058] See also Figure 1-Figure 2 The present invention provides a technical solution: a unified data management method based on machine learning, the method comprising:

[0059] Step S100: In the managed data cloud platform, manually review the resource data that has been through the automated classification system, collect and analyze the number of errors in the resource data reviewed as errors, and obtain a number of time intervals;

[0060] Wherein, step S100 includes:

[0061] Step S101: manually review the i-th resource data, collect the number of errors that are reviewed as errors, and plot a curve chart of the number of errors in the i-th resource data;

[0062] Step S102: Calculate the curve graph to obtain the change period of the i-th resource data, aggregate and calculate the change periods of various resource data in the managed data cloud platform, and obtain the average A and median B of the resource data change periods;

[0063] Step S103: Calculate the time interval for auditing resource data based on the mean and median, and calculate the time interval formula: G n =A+n×B, where G n It represents the nth time interval, n represents the weight of the time interval, and n is randomly selected within the specified range. The random value is input into the time interval formula to obtain several time intervals.

[0064] For example Figure 3 As shown, f1, f2, and f3 are respectively the positive and negative change points of the slope value of the error quantity curve of the first type of resource data. The first time period when the positive and negative slope value changes twice is the time period from the origin to f2. The first time period is the first change cycle. All change cycles are calculated to obtain the average change cycle, and the change cycle of the first type of resource data is the average change cycle.

[0065] Step S200: establishing an optimal prediction model for the number of errors in various resource data based on a number of time intervals;

[0066] Wherein, step S200 includes:

[0067] Step S201: G n As the time interval, the number of errors in the i-th resource data is collected, and a G n For the time interval, the error number prediction model of the i-th resource data is tested, and the prediction model is obtained by G n For the time interval, the error value of the error quantity prediction model of the i-th resource data will be expressed as G n For the time interval, the error values ​​of the error quantity prediction models of various resource data are added together to obtain the G n is the total error value of the time interval;

[0068] Step S202: Retrieve several time intervals, summarize the total error values ​​of each time interval, sort them from small to large, take the time interval corresponding to the minimum value as the optimal time interval, and use the error quantity prediction model of various resource data corresponding to the optimal time interval as the optimal error quantity prediction model of various resource data.

[0069] Step S300: Based on a threshold of the total time spent by staff members reviewing error data, obtain a plurality of combinations of error quantities for various resource data and a set of time spent on the plurality of combinations of error quantities;

[0070] Wherein, step S300 includes:

[0071] Step S301: Collect the time spent by staff on manual review of resource data, and calculate the average time spent by staff on manual review of resource data as T p , according to the average time, set a threshold T of the total time spent by staff members reviewing erroneous data;

[0072] Step S302: Collect the average time spent by staff on reviewing the error data of the i-th resource data, and calculate the total time spent on reviewing the number of errors in various resource data. According to the total time spent on review formula: T'=T1×S1+T2×S2+...+T i ×S i , where T' represents the total time spent on the review, T1, T2, ..., T i They are the average time spent on error data review for the 1st, 2nd, ..., i-th resource data, S1, S2, ..., S i They are represented as the number of errors in the 1st, 2nd, ..., i-th resource data respectively;

[0073] Step S303: The jth error quantity combination that satisfies the total time spent on reviewing is lower than the total time spent on reviewing error data is S 1j 、S 2j ,...,S ij , where S 1j 、S 2j ,...,S ij They are respectively represented as the number of errors of the 1st, 2nd, ..., i-th resource data in the j-th error number combination;

[0074] Step S304: S ij Input into the optimal prediction model of the number of errors of the i-th resource data, and the number of errors of the i-th resource data reaches S ij The time taken is to summarize the time taken for the number of errors of various resource data to reach the number of errors of various resource data in the jth error number combination to obtain the time set taken for the jth error number combination.

[0075] Step S400: Calculate the optimal level of the time spent based on a set of time spent for several combinations of error quantities, and update and analyze the automated classification system based on the range of the optimal level.

[0076] Wherein, step S400 includes:

[0077] Step S401: Retrieve a set of time spent for various combinations of error numbers, input the set of time spent into a K-means clustering algorithm, calculate the sum of squared errors of the K-means clustering, observe the change of the sum of squared errors using the elbow rule, and determine that the time spent has r levels;

[0078] Step S402: comparing the number of times spent in each level, and taking the level with the largest number of times spent as the optimal level;

[0079] Step S403: When the time between the last update time and the present time of the automated classification system reaches the minimum value of the optimal level, an update reminder is issued to the automated classification system. When the time between the last update time and the present time of the automated classification system reaches the maximum value of the optimal level, the automated classification system is shut down and an alarm is issued.

[0080] In order to better implement the above method, a unified data management system based on machine learning is proposed. The system includes a time interval module, a prediction model module, a quantity combination module, and an update duration module.

[0081] Time interval module: In the Naguan Data Cloud Platform, resource data that has been automatically classified by the system is manually reviewed. The number of errors identified in the review results for various resource data is collected and analyzed to obtain several time intervals.

[0082] Wherein, the time interval module includes a collection time interval unit and a calculation time interval unit;

[0083] Collection time interval unit: manually review various resource data, collect the number of errors that are reviewed as errors, and draw a curve chart of the number of errors in various resource data;

[0084] Calculation time interval unit: Analyze the curve graph to obtain the change cycle of various resource data curve graphs, perform weighted calculation, and obtain several time intervals.

[0085] Prediction model module: establishes the optimal prediction model for the number of errors in various resource data based on several time intervals;

[0086] Among them, the prediction model module includes a prediction model establishment unit and an optimal prediction model analysis unit;

[0087] Establishing a prediction model unit: collecting the number of errors in various resource data at different time intervals, and establishing a prediction model for the number of errors in various resource data collected at different time intervals;

[0088] Analyze the optimal prediction model unit: train the prediction model to obtain the total error value of the error quantity prediction model of various resource data collected at different time intervals, and take the prediction model with the smallest total error value as the optimal prediction model.

[0089] Quantity combination module: Based on the threshold of the total time spent by staff members reviewing error data, several error quantity combinations of various resource data and the time spent on several error quantity combinations are obtained;

[0090] Wherein, the quantity combination module includes an error time analysis unit and an error quantity combination unit;

[0091] Error time analysis unit: The time spent by staff on manual review of resource data is collected to obtain the average time. The average time is weighted and calculated to obtain the threshold of the total time spent by staff on reviewing error data.

[0092] Error quantity combination unit: Calculate based on the threshold of the total time spent by staff on reviewing error data and the average time spent on reviewing error data of various resource data to obtain several combinations of error quantities of various resource data. Summarize the time it takes for the error quantity of various resource data to reach the error quantity of various resource data in various error quantity combinations to obtain a set of time spent on various error quantity combinations.

[0093] Update duration module: Calculates the optimal level of the time spent based on a set of time spent for several combinations of error numbers, and performs update analysis on the automated classification system based on the range of the optimal level;

[0094] Among them, the update duration module includes an update duration analysis unit and an update reminder unit;

[0095] Analyze and update the duration unit: obtain various levels of the duration according to the duration of various error number combinations, and take the duration level with the largest number of errors as the optimal level;

[0096] Update reminder unit: When the time between the last update time and the present time of the automated classification system reaches the minimum value of the optimal level, an update reminder will be issued to the automated classification system. When the time between the last update time and the present time of the automated classification system reaches the maximum value of the optimal level, the automated classification system will be shut down and an alarm will be issued.

[0097] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0098] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A unified data management method based on machine learning, characterized by: Methods include: Step S100: In the data cloud platform, manually review the resource data that has passed the automated classification system, collect and analyze the number of errors in the resource data review results to obtain a number of time intervals; In S100, the specific steps are as follows: Step S101: manually review the i-th resource data, collect the number of errors whose review results are errors, and draw a curve chart of the number of errors in the i-th resource data; Step S102: Calculate the curve graph to obtain the change period of the i-th resource data, summarize and calculate the change periods of various resource data in the managed data cloud platform, and obtain the average A and median B of the resource data change period; Step S103: Calculate the time interval for auditing resource data based on the mean and median, and use the time interval formula: G n =A+n×B, where G n It is represented as the nth time interval, n is represented as the weight of the time interval, and n is randomly selected within the specified range, and the random value is input into the time interval formula to obtain a number of time intervals; Step S200: establishing an optimal prediction model for the number of errors in various resource data according to a number of time intervals; In S200, the specific steps are as follows: Step S201: G n As the time interval, the number of errors in the i-th resource data is collected to establish a G n is the time interval, the error number prediction model of the i-th resource data is tested, and the prediction model is obtained by G n is the time interval, and the error value of the error quantity prediction model of the i-th resource data will be expressed as G n The error values ​​of the error quantity prediction models of various resource data are added together to obtain the G n is the total error value of the time interval; Step S202: Retrieve a number of time intervals, summarize the total error values ​​of each time interval, sort them from small to large, take the time interval corresponding to the minimum value as the optimal time interval, and take the error quantity prediction model of various resource data corresponding to the optimal time interval as the error quantity optimal prediction model of various resource data; Step S300: according to the threshold of the total time spent by the staff to review the error data, obtain several combinations of error quantities of various resource data and sets of time spent on the several combinations of error quantities; Step S400: Calculate the optimal level of the time spent according to the time spent for several combinations of error quantities, and update and analyze the automatic classification system according to the range of the optimal level.

2. The unified data management method based on machine learning according to claim 1 is characterized in that: Step S300 is a process of obtaining a plurality of combinations of error quantities of various resource data and a set of time durations spent on the plurality of combinations of error quantities, including the following steps: Step S301: Collect the time spent by the staff on manual review of resource data, and calculate the average time spent by the staff on manual review of resource data as T p , according to the average time, set a threshold T of the total time spent by the staff in reviewing the erroneous data; Step S302: Collect the average time spent by the staff on reviewing the erroneous data of the i-th resource data, and calculate the total time spent on reviewing the number of errors in various resource data. According to the total time spent on review formula: T'=T1×S1+T2×S2+...+T i ×S i , where T' represents the total time spent on the review, T1, T2, ..., T i are the average time spent on error data review for the 1st, 2nd, ..., i-th resource data, S1, S2, ..., S i They are represented as the number of errors in the 1st, 2nd, ..., i-th resource data respectively; Step S303: The jth error quantity combination that satisfies the total time spent on reviewing is lower than the total time spent on reviewing the error data is S 1j , S 2j ,...,S ij , where S 1j , S 2j ,...,S ij They are respectively represented as the number of errors of the 1st, 2nd, ..., i-th resource data in the j-th error number combination; Step S304: S ij Input into the optimal prediction model of the number of errors in the i-th resource data, and the number of errors in the i-th resource data reaches S ij The time taken is to summarize the time taken for the number of errors of various resource data to reach the number of errors of various resource data in the jth error number combination, and obtain the time taken set of the jth error number combination.

3. The unified data management method based on machine learning according to claim 2 is characterized in that: The process of obtaining the update interval duration in step S400 includes the following steps: Step S401: Retrieve a set of time spent for various combinations of error quantities, input the set of time spent into the K-means clustering algorithm, calculate the sum of square errors of the K-means clustering, observe the change of the sum of square errors through the elbow rule, and obtain that the time spent has r levels; Step S402: comparing the quantities contained in each level of the time spent, and taking the level of the time spent with the largest quantity as the optimal level; Step S403: When the time between the last update time and the present time of the automated classification system reaches the minimum value of the optimal level, an update reminder is issued to the automated classification system; when the time between the last update time and the present time of the automated classification system reaches the maximum value of the optimal level, the automated classification system is shut down and an alarm is issued.

4. A unified data management system based on machine learning, used to implement a unified data management method based on machine learning as described in any one of claims 1 to 3, characterized in that: Time interval module, prediction model module, quantity combination module and update duration module; The time interval module: in the data cloud platform, manually review the resource data that has passed the automated classification system, collect and analyze the number of errors in the review results of various resource data, and obtain a number of time intervals; The prediction model module: establishes an optimal prediction model for the number of errors in various resource data according to a number of time intervals; The quantity combination module: obtains several combinations of error quantities of various resource data and sets of time spent on the several combinations of error quantities according to a threshold of the total time spent by the staff in reviewing the error data; The update time module calculates the optimal level of the time spent according to the time spent for several combinations of error quantities, and performs update analysis on the automatic classification system according to the range of the optimal level.

5. The unified data management system based on machine learning according to claim 4 is characterized in that: The time interval module includes a collection time interval unit and a calculation time interval unit; The collection time interval unit: manually review various resource data, collect the number of errors whose review results are errors, and draw a curve chart of the number of errors in various resource data; The time interval calculation unit analyzes the curve graph to obtain the change period of various resource data curve graphs, performs weighted calculation, and obtains a plurality of time intervals.

6. The unified data management system based on machine learning according to claim 4, characterized in that: The prediction model module includes a prediction model establishment unit and an optimal prediction model analysis unit; The prediction model establishment unit collects the number of errors of various resource data at different time intervals and establishes a prediction model for the number of errors of various resource data collected at different time intervals; The optimal prediction model analysis unit trains the prediction model to obtain the total error value of the error quantity prediction model of various resource data collected at different time intervals, and takes the prediction model with the smallest total error value as the optimal prediction model.

7. The unified data management system based on machine learning according to claim 4 is characterized in that: The quantity combination module includes an error time analysis unit and an error quantity combination unit; The error analysis time unit: collects the time spent by the staff on manually reviewing the resource data to obtain the average time, performs weighted calculation on the average time, and obtains the threshold of the total time spent by the staff on reviewing the error data; The error quantity combination unit: calculates according to the threshold of the total time spent by the staff on reviewing the error data and the average time spent on reviewing the error data of various resource data, obtains several error quantity combinations of various resource data, summarizes the time spent for the error quantity of various resource data to reach the error quantity of various resource data in various error quantity combinations, and obtains a set of time spent for various error quantity combinations.

8. The unified data management system based on machine learning according to claim 4, characterized in that: The update duration module includes an update duration analysis unit and an update reminder unit; The analyzing and updating time unit: obtaining various levels of the time spent according to the time spent by various error quantity combinations, and taking the time spent level with the largest number in the spent level as the optimal level; The update reminder unit sends an update reminder to the automated classification system when the time between the last update time and the present time reaches the minimum value of the optimal level; and shuts down the automated classification system and sends an alarm when the time between the last update time and the present time reaches the maximum value of the optimal level.

Images