Big data statistical optimization system based on reinforcement learning

Abstract

The invention discloses a reinforcement learning-based big data statistical optimization system. The system comprises a data acquisition module and a partition storage module, the partition storage module is provided with a plurality of area blocks according to the obtained historical learning item data, each area block corresponds to one learning item and stores all related knowledge type data corresponding to the historical learning item, and the data acquisition module is provided with an independent temporary storage space and is specifically used for storing the historical learning item data. Based on the comparison of the data size and the residual storage capacity and in combination with the judgment criterion of the grade value, the sub-region blocks to be allocated are determined, the capacity in the region blocks can be utilized to the maximum degree to carry out the self-adaptive storage space adjustment strategy, and meanwhile after all data are determined, integrated data statistics is carried out according to the data, so that the allocation efficiency is improved. And performing secondary execution correction on the obtained fitting curve value, and performing optimization statistical recommendation after correction, thereby facilitating statistical collection and optimization on learning data.

Description

Technical Field

[0001] The present invention relates to the field of data collection and processing technology in the field of Internet of Things, and specifically to a big data statistical optimization system based on reinforcement learning. Background Art

[0002] In today's information age, big data has become an indispensable strategic resource for all walks of life. However, with the explosive growth of data volume and the diversification of data types, traditional statistical and optimization methods face challenges such as low computational efficiency, poor model adaptability, and difficulty in capturing complex relationships between data when processing big data. Therefore, how to efficiently process and optimize big data to extract valuable information and serve practical applications has become a problem that needs to be solved urgently.

[0003] As an advanced machine learning technology, reinforcement learning has shown its powerful ability in solving decision-making problems in complex environments with its unique "trial and error-learning-optimization" mechanism. By converting the big data statistical optimization task into a reinforcement learning problem, that is, treating the data processing process as the process of interaction between the intelligent agent and the environment, where the data is the environmental state, the data processing strategy is the action of the intelligent agent, and the statistical indicators or business goals are the reward signals, the automatic optimization of the big data processing strategy can be achieved.

[0004] The existing learning optimization system is relatively scattered, and the optimization information used in each region is counted separately, which is inconvenient to manage. At the same time, the statistical optimization method of learning big data in each region is single and the content is monotonous, resulting in low accuracy of big data information on learning. Therefore, it is not convenient for staff to summarize and conclude big data information regularly. Summary of the invention

[0005] The purpose of the present invention is to provide a big data statistical optimization system based on reinforcement learning to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: A big data statistical optimization system based on reinforcement learning, comprising: a data acquisition module, a partition storage module;

[0007] The partition storage module is provided with a plurality of area blocks according to the acquired historical learning project data, each area block corresponds to a learning project and stores all relevant knowledge type data corresponding to the historical learning project;

[0008] The data acquisition module is provided with an independent temporary storage space, which is specifically used for:

[0009] S1, real-time collection of newly generated published implementation data of different learning projects and temporary storage to a temporary storage space, each learning project includes at least one overall viewing reference data, each overall viewing reference data includes project experimental data, paper data, book data and project report data;

[0010] S2, inputting all the coordinated viewing reference data of different learning type projects into the pre-trained level recognition model to obtain the level value of each learning difficulty in the coordinated viewing reference data;

[0011] S3, presetting a time period, and periodically storing the coordinated viewing reference data of all different learning projects within the time period into the partition storage module, specifically including:

[0012] Based on each learning project in the time period, determine the target area block in the partition storage module;

[0013] Determine each learning project within the time period as a learning project to be stored and transfer it to the corresponding target area block;

[0014] S4. Preset multiple different learning partitions according to the preset algorithm based on the current social development environment, periodically replace the learning data in different target area blocks, and make statistical planning data values ​​based on the replaced learning data time data and content data;

[0015] S5. Feed the statistical planning data value into the linear regression model, optimize the model parameters by minimizing the residual sum of squares between the predicted value and the true value, and obtain the best fitting curve. Then, optimize and correct step S2 by using the obtained best fitting curve value. After integration, fit the corresponding level of learning projects based on the current input keywords and feature values.

[0016] Preferably, in S3, based on each different learning project in the time period, determining the target area block in the partition storage module specifically includes:

[0017] B1. Based on each different learning project in the time period, determine whether there is a historical learning project record with the same learning type project in the area block of the partition storage module. If so, determine the corresponding area block as the target area block, and perform a traversal and loop comparison on the data of the historical learning project record and the data of the learning project to obtain a difference value. When a difference value exists, open up a separate integrated storage module and perform the integrated comparison step. When no difference value exists, replace the historical learning project record data and update the record date.

[0018] B2. If it does not exist, the similarity between the learning type project and the learning type project in each region block in the partition storage module is calculated according to the following formula, and the region block with the smallest similarity value is used as the target region block:

[0019]

[0020] Among them, S ij is the similarity value between the current learning project i and the historical learning project j, Q is the number of the same coordinated viewing reference data in the current learning project i and the historical learning project j, and W q is the grade value of the qth reference data viewed by the same overall plan, M iq 、M jq are the project experimental data and project data report of the qth coordinated viewing reference data of the current learning project i and the historical learning project j, respectively. iq 、N jq are the paper data and project-based book data of the qth coordinated viewing reference data of the current learning project i and the historical learning project j, respectively. α and β are the weight factors for adjusting the contribution of the difference in the number of experimental data and project-based books to the similarity value.

[0021] Preferably, the step of performing integration comparison in step B1 specifically includes:

[0022] B11. When the partition storage module finds that there are historical learning project records that are the same as the learning projects within the time period, but there are difference values, first identify the specific contents of these differences, including but not limited to the newly added overall viewing reference data, updated project experiment data, paper data, book data and project report data;

[0023] B12. Allocate space for the newly added data in a separate integrated storage module and retain the data recorded in the original historical learning project;

[0024] B13. Conduct a detailed comparative analysis of new and old data, including content review of new and updated data, assessing their impact on the overall learning project, and calculating the correlation and complementarity between new and existing data. Based on the correlation and complementarity, this paper provides multiple correlation and complementarity analysis results to obtain integrated data;

[0025] B14. Based on the results of the comparative analysis, the integrated data is reclassified and labeled to ensure that it accurately reflects the latest status and characteristics of the current learning project.

[0026] B15. Store the integrated data in the corresponding target area block, update the record date, and record the detailed process and results of this integration comparison, waiting for the next round of optimization traversal.

[0027] Preferably, the preset algorithm in S4 is specifically:

[0028] Among them, G1 is the current social environment information comment data benchmark value, are the average values ​​of the learning time data, learning content data, and learning development direction data of all time nodes within the preset time period, [X1, X2], [B1, B2], and [V1, V2] are the normal ranges of the preset learning time range, the normal ranges of the learning development direction, and the normal ranges of the learning content data, respectively. e1, e2, and e3 are the weight factors corresponding to the learning time data values, learning content data, and learning development direction data values, respectively.

[0029] Preferably, the partition storage module is also used for:

[0030] S201, based on each area block, according to the number of coordinated viewing reference data in the corresponding historical learning type project, the area block is divided into a corresponding number of sub-area blocks, each sub-area block corresponds to a learning project type, and all coordinated viewing reference data types in the historical learning project type are stored in the corresponding sub-area blocks;

[0031] S202, when receiving the learning type items to be stored transmitted by the data acquisition module, according to the level values ​​of all the coordinated viewing reference data, the coordinated viewing reference data of each coordinated viewing reference data is stored in the sub-region block of the corresponding level value in turn.

[0032] Preferably, S301, determining whether there is a data volume of the coordinated viewing reference data in the cultural tourism project to be stored that is greater than the remaining storage capacity of the corresponding sub-region block;

[0033] S302, if it exists, the coordinated viewing reference data with a data volume greater than the remaining storage capacity of the corresponding sub-regional block is determined as the first type of coordinated viewing reference data: when the level value of the first type of coordinated viewing reference data is greater than the preset level threshold, all sub-regional blocks with remaining storage capacity and corresponding coordinated viewing reference data level values ​​less than the preset level threshold are determined as sub-regional blocks to be allocated, and step S303 is executed; otherwise, the coordinated viewing reference data of the invalid learning type project in the sub-regional block corresponding to the first type of coordinated viewing reference data is deleted;

[0034] S303: Split the remaining storage capacity in the sub-region block to be allocated, and expand it to the sub-region block corresponding to the first type of coordinated viewing reference data.

[0035] Preferably, in S5, the learning optimization value is calculated according to the existing linear regression model, and the overall viewing reference data is used as a quantitative indicator, and the calculation formula is: Estimated View Count = β0 + β1 × DifficultyLevel + β2 × Publish Date (DaysSinceEpoch) +

[0036] β3×ViewCount+β4×UserFeedbackScore+β5×SocialMediaMentions.

[0037] Among them, Estimated View Count is the dependent variable to be predicted, that is, the target variable;

[0038] β0 is the bias term or intercept, which represents the expected value of the dependent variable when all independent variables are 0;

[0039] β1, β2, β3, β4, and β5 are regression coefficients or weights, corresponding to the degree of influence of each independent variable.

[0040] Preferably, after step S5 is completed, the following steps are performed, specifically:

[0041] X1. Make predictions based on the application data in the current learning project data partition and give corresponding data reference values ​​for the corresponding optimized overall viewing reference data.

[0042] Preferably, in X1, after the corresponding data reference value of the corresponding optimized coordinated viewing reference data is given, the corresponding learning difficulty level value is matched according to the reference value, and the data reference value and the learning difficulty level value are used as anchor points to re-retrieve the data, and S1 is returned again after the data retrieval.

[0043] Preferably, in the process of re-searching data with the data reference value and the learning difficulty level value as anchor points, a search threshold number is set, and the search threshold number is 3 years before and after the current time value.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] By setting temporary storage space and periodic transfer storage strategy, the real-time processing pressure is reduced and the system processing efficiency is improved. In addition, periodic data transfer and storage can provide sufficient and reliable reference data for learning project coordination within a reliable time period to calculate the target area block, thereby improving the accuracy and reliability of matching. By calculating the similarity between the learning type project and the existing area block, it can be stored in the most similar area block, without the need to create a new area block separately, saving system resources and avoiding chaos within the area block. By dividing each area block into sub-area blocks, the accurate storage and management of reference data for learning project coordination is achieved, and the accuracy and efficiency of data storage are improved. Based on the comparison of data size and remaining storage capacity, combined with the evaluation criteria of the grade value, the sub-area block to be allocated is determined, which can maximize the use of the capacity within the area block to perform adaptive storage space adjustment strategies. At the same time, after determining each data, the data is integrated and statistically analyzed according to the data, and the fitting curve value is obtained for secondary execution correction, and the optimization statistical recommendation is performed based on the correction, which is convenient for statistical collection and optimization of learning data. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example 1: Please refer to Figure 1 , the present invention provides a technical solution:

[0049] A big data statistical optimization system based on reinforcement learning, including: data acquisition module, partition storage module;

[0050] The partition storage module is provided with a plurality of area blocks according to the acquired historical learning project data, each area block corresponds to a learning project and stores all relevant knowledge type data corresponding to the historical learning project;

[0051] The data acquisition module is provided with an independent temporary storage space, which is specifically used for:

[0052] S1, real-time collection of newly generated published implementation data of different learning projects and temporary storage to a temporary storage space, each learning project includes at least one overall viewing reference data, each overall viewing reference data includes project experimental data, paper data, book data and project report data;

[0053] S2, inputting all the coordinated viewing reference data of different learning type projects into the pre-trained level recognition model to obtain the level value of each learning difficulty in the coordinated viewing reference data;

[0054] S3, presetting a time period, and periodically storing the coordinated viewing reference data of all different learning projects within the time period into the partition storage module, specifically including:

[0055] Based on each learning project in the time period, determine the target area block in the partition storage module;

[0056] Determine each learning project within the time period as a learning project to be stored and transfer it to the corresponding target area block;

[0057] S4. Preset multiple different learning partitions according to the preset algorithm based on the current social development environment, periodically replace the learning data in different target area blocks, and make statistical planning data values ​​based on the replaced learning data time data and content data;

[0058] S5. Feed the statistical planning data value into the linear regression model, optimize the model parameters by minimizing the residual sum of squares between the predicted value and the true value, and obtain the best fitting curve. Then, optimize and correct step S2 by using the obtained best fitting curve value. After integration, fit the corresponding level of learning projects based on the current input keywords and feature values.

[0059] Specifically, in S3, based on each different learning project in the time period, determining the target area block in the partition storage module specifically includes:

[0060] B1. Based on each different learning project in the time period, determine whether there is a historical learning project record with the same learning type project in the area block of the partition storage module. If so, determine the corresponding area block as the target area block, and perform a traversal and loop comparison on the data of the historical learning project record and the data of the learning project to obtain a difference value. When a difference value exists, open up a separate integrated storage module and perform the integrated comparison step. When no difference value exists, replace the historical learning project record data and update the record date.

[0061] B2. If it does not exist, the similarity between the learning type project and the learning type project in each region block in the partition storage module is calculated according to the following formula, and the region block with the smallest similarity value is used as the target region block:

[0062]

[0063] Among them, S ij is the similarity value between the current learning project i and the historical learning project j, Q is the number of the same coordinated viewing reference data in the current learning project i and the historical learning project j, and W q is the grade value of the qth reference data viewed by the same overall plan, M iq 、M jq are the project experimental data and project data report of the qth coordinated viewing reference data of the current learning project i and the historical learning project j, respectively. iq 、N jq are the paper data and project-based book data of the qth coordinated viewing reference data of the current learning project i and the historical learning project j, respectively. α and β are the weight factors for adjusting the contribution of the difference in the number of experimental data and project-based books to the similarity value.

[0064] Specifically, the step of performing integration comparison in step B1 specifically includes:

[0065] B11. When the partition storage module finds that there are historical learning project records that are the same as the learning projects within the time period, but there are difference values, first identify the specific contents of these differences, including but not limited to the newly added overall viewing reference data, updated project experiment data, paper data, book data and project report data;

[0066] B12. Allocate space for the newly added data in a separate integrated storage module and retain the data recorded in the original historical learning project;

[0067] B13. Conduct a detailed comparative analysis of new and old data, including content review of new and updated data, assessing their impact on the overall learning project, and calculating the correlation and complementarity between new and existing data. Based on the correlation and complementarity, this paper provides multiple correlation and complementarity analysis results to obtain integrated data;

[0068] B14. Based on the results of the comparative analysis, the integrated data is reclassified and labeled to ensure that it accurately reflects the latest status and characteristics of the current learning project.

[0069] B15. Store the integrated data in the corresponding target area block, update the record date, and record the detailed process and results of this integration comparison, waiting for the next round of optimization traversal.

[0070] Specifically, the preset algorithm in S4 is:

[0071] Among them, G1 is the current social environment information comment data benchmark value, are the average values ​​of the learning time data, learning content data, and learning development direction data of all time nodes within the preset time period, [X1, X2], [B1, B2], and [V1, V2] are the normal ranges of the preset learning time range, the normal ranges of the learning development direction, and the normal ranges of the learning content data, respectively. e1, e2, and e3 are the weight factors corresponding to the learning time data values, learning content data, and learning development direction data values, respectively.

[0072] Embodiment 2: In some embodiments, the partition storage module is also used for:

[0073] S201, based on each area block, according to the number of coordinated viewing reference data in the corresponding historical learning type project, the area block is divided into a corresponding number of sub-area blocks, each sub-area block corresponds to a learning project type, and all coordinated viewing reference data types in the historical learning project type are stored in the corresponding sub-area blocks;

[0074] S202, when receiving the learning type items to be stored transmitted by the data acquisition module, according to the level values ​​of all the coordinated viewing reference data, the coordinated viewing reference data of each coordinated viewing reference data is stored in the sub-region block of the corresponding level value in turn.

[0075] Specifically, S301, determining whether there is a data volume of the coordinated viewing reference data in the cultural tourism project to be stored that is greater than the remaining storage capacity of the corresponding sub-region block;

[0076] S302, if it exists, the coordinated viewing reference data with a data volume greater than the remaining storage capacity of the corresponding sub-regional block is determined as the first type of coordinated viewing reference data: when the level value of the first type of coordinated viewing reference data is greater than the preset level threshold, all sub-regional blocks with remaining storage capacity and corresponding coordinated viewing reference data level values ​​less than the preset level threshold are determined as sub-regional blocks to be allocated, and step S303 is executed; otherwise, the coordinated viewing reference data of the invalid learning type project in the sub-regional block corresponding to the first type of coordinated viewing reference data is deleted;

[0077] S303: Split the remaining storage capacity in the sub-region block to be allocated, and expand it to the sub-region block corresponding to the first type of coordinated viewing reference data.

[0078] Specifically, in S5, the learning optimization value is calculated according to the existing linear regression model, and the overall viewing reference data is used as a quantitative indicator. The calculation formula is: Estimated View Cound = β0 + β1 × DifficultyLevel + β2 × Publish Date (DaysSinceEpoch) +

[0079] β3×ViewCount+β4×UserFeedbackScore+β5×SocialMediaMentions.

[0080] Among them, Estimated View Count is the dependent variable to be predicted, that is, the target variable;

[0081] β0 is the bias term or intercept, which represents the expected value of the dependent variable when all independent variables are 0;

[0082] β1, β2, β3, β4, and β5 are regression coefficients or weights, corresponding to the degree of influence of each independent variable.

[0083] Embodiment three:

[0084] Considering that after executing S5, the partitioning effect needs to be analyzed and predicted, the specific steps of analysis and prediction are as follows:

[0085] X1. Make predictions based on the application data in the current learning project data partition and give corresponding data reference values ​​for the corresponding optimized overall viewing reference data.

[0086] Specifically, in X1, after the corresponding data reference value of the corresponding optimized coordinated viewing reference data is given, the corresponding learning difficulty level value is matched according to the reference value, and the data reference value and the learning difficulty level value are used as anchor points to re-retrieve the data, and S1 is returned again after the data retrieval.

[0087] Specifically, during the data retrieval process using the data reference value and the learning difficulty level value as anchor points, a retrieval threshold number is set, and the retrieval threshold number is 3 years before and after the current time value.

[0088] According to the above technical solution, the working steps of this solution are summarized and sorted out: by setting temporary storage space and periodic transfer storage strategy, the real-time processing pressure is reduced, the system processing efficiency is improved, and the periodic data transfer and storage can provide sufficient and reliable time period for the reference data of learning project coordinated viewing to calculate the target area block, thereby improving the accuracy and reliability of matching; by calculating the similarity between the learning type project and the existing area block, it can be stored in the most similar area block, without the need to create a new area block separately, saving system resources and avoiding chaos within the area block; by dividing each area block into sub-area blocks, the accurate storage and management of the reference data of the coordinated viewing of the learning project is realized, the accuracy and efficiency of data storage are improved, and based on the comparison of the data volume and the remaining storage capacity, combined with the evaluation criteria of the grade value, the sub-area block to be allocated is determined, which can maximize the use of the capacity within the area block to perform the adaptive storage space adjustment strategy, and after determining the various data, the data statistics are integrated according to the data, the fitting curve value is obtained for secondary execution correction, and the optimization statistics recommendation is made according to the correction, which is convenient for statistical collection and optimization of learning data.

[0089] Parts not involved in the present invention are the same as the prior art or can be implemented by the prior art. Although the embodiments of the present invention have been shown and described, it is understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A big data statistical optimization system based on reinforcement learning, characterized in that: include: Data acquisition module, partition storage module; The partition storage module is provided with a plurality of area blocks according to the acquired historical learning project data, each area block corresponds to a learning project and stores all relevant knowledge type data corresponding to the historical learning project; The data acquisition module is provided with an independent temporary storage space, which is specifically used for: S1, real-time collection of newly generated published implementation data of different learning projects and temporary storage to a temporary storage space, each learning project includes at least one overall viewing reference data, each overall viewing reference data includes project experimental data, paper data, book data and project report data; S2, inputting all the coordinated viewing reference data of different learning type projects into the pre-trained level recognition model to obtain the level value of each learning difficulty in the coordinated viewing reference data; S3, presetting a time period, and periodically storing the coordinated viewing reference data of all different learning projects within the time period into the partition storage module, specifically including: Based on each learning project in the time period, determine the target area block in the partition storage module; Determine each learning project within the time period as a learning project to be stored and transfer it to the corresponding target area block; S4. Preset multiple different learning partitions according to the preset algorithm based on the current social development environment, periodically replace the learning data in different target area blocks, and make statistical planning data values ​​based on the replaced learning data time data and content data; S5. Feed the statistical planning data value into the linear regression model, optimize the model parameters by minimizing the residual sum of squares between the predicted value and the true value, and obtain the best fitting curve. Then, optimize and correct step S2 by using the obtained best fitting curve value. After integration, fit the corresponding level of learning projects based on the current input keywords and feature values.

2. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: In S3, based on each different learning project in the time period, a target area block is determined in the partition storage module, specifically including: B1. Based on each different learning project in the time period, determine whether there is a historical learning project record with the same learning type project in the area block of the partition storage module. If so, determine the corresponding area block as the target area block, and perform a traversal and loop comparison on the data of the historical learning project record and the data of the learning project to obtain a difference value. When a difference value exists, open up a separate integrated storage module and perform the integrated comparison step. When no difference value exists, replace the historical learning project record data and update the record date. B2. If it does not exist, the similarity between the learning type project and the learning type project in each region block in the partition storage module is calculated according to the following formula, and the region block with the smallest similarity value is used as the target region block: Among them, S ij is the similarity value between the current learning project i and the historical learning project j, Q is the number of the same coordinated viewing reference data in the current learning project i and the historical learning project j, and W q is the grade value of the qth reference data viewed by the same overall plan, M iq 、M jq are the project experimental data and project data report of the qth coordinated viewing reference data of the current learning project i and the historical learning project j, respectively. iq 、N jq are the paper data and project-based book data of the qth coordinated viewing reference data of the current learning project i and the historical learning project j, respectively. α and β are the weight factors for adjusting the contribution of the difference in the number of experimental data and project-based books to the similarity value.

3. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: The step of performing integration comparison in B1 specifically includes: B11. When the partition storage module finds that there are historical learning project records that are the same as the learning projects within the time period, but there are difference values, first identify the specific contents of these differences, including but not limited to the newly added overall viewing reference data, updated project experiment data, paper data, book data and project report data; B12. Allocate space for the newly added data in a separate integrated storage module and retain the data recorded in the original historical learning project; B13. Conduct a detailed comparative analysis of new and old data, including content review of new and updated data, assessing their impact on the overall learning project, and calculating the correlation and complementarity between new and existing data. Based on the correlation and complementarity, this paper provides multiple correlation and complementarity analysis results to obtain integrated data; B14. Based on the results of the comparative analysis, the integrated data is reclassified and labeled to ensure that it accurately reflects the latest status and characteristics of the current learning project. B15. Store the integrated data in the corresponding target area block, update the record date, and record the detailed process and results of this integration comparison, waiting for the next round of optimization traversal.

4. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: The preset algorithm in S4 is specifically: Among them, G1 is the current social environment information comment data benchmark value, are the average values ​​of the learning time data, learning content data, and learning development direction data of all time nodes within the preset time period, [X1, X2], [B1, B2], and [V1, V2] are the normal ranges of the preset learning time range, the normal ranges of the learning development direction, and the normal ranges of the learning content data, respectively. e1, e2, and e3 are the weight factors corresponding to the learning time data values, learning content data, and learning development direction data values, respectively.

5. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: The partition storage module is also used for: S201, based on each area block, according to the number of coordinated viewing reference data in the corresponding historical learning type project, the area block is divided into a corresponding number of sub-area blocks, each sub-area block corresponds to a learning project type, and all coordinated viewing reference data types in the historical learning project type are stored in the corresponding sub-area blocks; S202, when receiving the learning type items to be stored transmitted by the data acquisition module, according to the level values ​​of all the coordinated viewing reference data, the coordinated viewing reference data of each coordinated viewing reference data is stored in the sub-region block of the corresponding level value in turn.

6. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: S301, determining whether there is a data volume of the coordinated viewing reference data in the cultural tourism project to be stored that is greater than the remaining storage capacity of the corresponding sub-region block; S302, if it exists, the coordinated viewing reference data with a data volume greater than the remaining storage capacity of the corresponding sub-regional block is determined as the first type of coordinated viewing reference data: when the level value of the first type of coordinated viewing reference data is greater than the preset level threshold, all sub-regional blocks with remaining storage capacity and corresponding coordinated viewing reference data level values ​​less than the preset level threshold are determined as sub-regional blocks to be allocated, and step S303 is executed; otherwise, the coordinated viewing reference data of the invalid learning type project in the sub-regional block corresponding to the first type of coordinated viewing reference data is deleted; S303: Split the remaining storage capacity in the sub-region block to be allocated, and expand it to the sub-region block corresponding to the first type of coordinated viewing reference data.

7. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: In S5, the learning optimization value is calculated according to the existing linear regression model, and the overall viewing reference data is used as the quantitative indicator. The calculation formula is: Estimated View Count = β0 + β1 × DifficultyLevel + β2 × Publish Date (DaysSinceEpoch) + β3 × ViewCount + β4 × UserFeedbackScore + β5 × SocialMediaMentions. Among them, Estimated View Count is the dependent variable to be predicted, that is, the target variable; β0 is the bias term or intercept, which represents the expected value of the dependent variable when all independent variables are 0; β1, β2, β3, β4, and β5 are regression coefficients or weights, corresponding to the degree of influence of each independent variable.

8. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: After step S5 is completed, the following steps are performed, specifically: X1. Make predictions based on the application data in the current learning project data partition and give corresponding data reference values ​​for the corresponding optimized overall viewing reference data.

9. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: In X1, after the corresponding data reference value of the corresponding optimized coordinated viewing reference data is given, the corresponding learning difficulty level value is matched according to the reference value, and the data reference value and the learning difficulty level value are used as anchor points to re-retrieve the data, and S1 is returned again after the data retrieval.

10. The big data statistical optimization system based on reinforcement learning according to claim 1, characterized in that: In the process of re-retrieval of data with the data reference value and the learning difficulty level value as anchor points, a retrieval threshold number is set, and the retrieval threshold number is 3 years before and after the current time value.

Images