A multi-dimensional data intelligent management method for regional security management software

Abstract

The application relates to the technical field of data processing, and discloses a multi-dimensional data intelligent management method for regional safety management software, which comprises the following steps: obtaining a left singular matrix, a right singular matrix, a diagonal line matrix and eigenvectors of the left singular matrix and the right singular matrix of a two-dimensional driving data matrix; obtaining a first kind of reservation degree, which is divided into time periods; obtaining an adjusted voting value and marking important time periods; obtaining an influence weight value, determining a first eigenvector, obtaining a reconstructed two-dimensional driving data matrix of the first eigenvector, obtaining a reconstructed influence weight value according to the reconstructed two-dimensional driving data matrix, and then obtaining a feature weight of the first eigenvector; and determining data that needs to be reserved and stored according to the feature weight, so as to realize multi-dimensional data intelligent management for the regional safety management software. The application aims to solve the problem of low multi-dimensional data intelligent management efficiency caused by excessive redundant data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to a multi-dimensional data intelligent management method for regional security management software. Background Technology

[0002] Area safety management is a complex and crucial issue, particularly in the realm of vehicle movement. During vehicle operation, countless sensors generate massive amounts of multi-dimensional data daily, including vehicle position, speed, direction, and environmental conditions. This sensor data is invaluable for understanding traffic conditions, predicting potential risks during vehicle movement, and optimizing routes. However, processing and storing this large-scale, high-dimensional data presents a significant challenge. Because the data used in area safety management is multi-dimensional, and each dimension is time-series data, the volume of sensor data acquired is enormous, requiring compression before storage in a database. A common approach is to use the SVD (Simultaneous Vector Decomposition) algorithm to decompose the two-dimensional driving data matrix composed of the acquired sensor data, and then store the decomposition results. This method can capture more effective information for area safety warnings while significantly reducing data storage.

[0003] However, the decomposition process still results in a large amount of redundant data. Furthermore, not all data contains valuable information that needs to be retained and analyzed during the compression process; in other words, there is a significant amount of redundant data. Therefore, directly compressing all data is often detrimental to data management and analysis, leading to wasted resources and reduced efficiency in multidimensional data intelligent management. Summary of the Invention

[0004] This invention provides a multi-dimensional data intelligent management method for regional security management software to solve the problem of low efficiency in multi-dimensional data intelligent management caused by excessive redundant data. The specific technical solution adopted is as follows: One embodiment of the present invention provides a multi-dimensional data intelligent management method for regional security management software, the method comprising the following steps: Collect vehicle driving status data, establish a two-dimensional driving data matrix, obtain the left singular matrix, right singular matrix, diagonal matrix of the two-dimensional driving data matrix, as well as the eigenvectors of the left singular matrix and right singular matrix, and obtain the number of system warnings and the time of each system warning; The driving data vector is obtained from the two-dimensional driving data matrix. The target time point is determined based on the driving data vector. The slope difference of the driving condition data of the vehicle is obtained. The first category is filtered based on the slope difference of the driving condition data of the vehicle. The retention degree of the first category is obtained. Each row of data in the two-dimensional driving data matrix is ​​divided into multiple time periods. Obtain the voting value within a preset window scale within the time period of the target time point; determine the weight value of the baseline voting value within the preset window scale; perform a weighted average of the baseline voting values ​​based on the weight values ​​of the baseline voting values ​​to obtain the baseline voting value of the target time point; determine the mean of the first cosine similarity at the first time point; determine the mean of the second cosine similarity at the target time point; obtain the adjusted voting value based on the baseline voting value of the target time point, the mean of the first cosine similarity at the first time point, and the mean of the second cosine similarity; obtain the voting distance of the vehicle driving status data based on the adjusted voting value; and mark important time periods based on the voting distance. The system obtains the influence weight value of driving status data for each type of vehicle, determines the first feature vector, obtains the reconstructed two-dimensional driving data matrix of the first feature vector, obtains the reconstruction influence weight value based on the reconstructed two-dimensional driving data matrix, obtains the feature weight of the first feature vector based on the influence weight value of driving status data for the same type of vehicle and the reconstruction influence weight value, and determines the data to be retained and stored based on the feature weight, thereby realizing multi-dimensional intelligent data management for regional safety management software.

[0005] Furthermore, the specific method for obtaining driving data vectors based on a two-dimensional driving data matrix and determining the target time point based on the driving data vectors includes: Each column in the two-dimensional driving data matrix is ​​treated as a driving data vector. Each driving data vector is taken as the driving data vector to be analyzed, and the data acquisition time corresponding to the data vector to be analyzed is recorded as the target time point.

[0006] Furthermore, the specific method for obtaining the slope difference of the vehicle's driving status data includes: The vehicle's driving status data at two data collection times after the target time point will be recorded as the adjacent driving status data at the target time point; The absolute value of the difference between the driving condition data at the next moment and the current moment is recorded as the driving condition data slope at the current moment. The driving condition data slopes at the target time point and the two data collection moments after the target time point are obtained. The average of the absolute values ​​of the differences in the slopes of the driving condition data at the target time point and at two subsequent data collection times is recorded as the slope difference of the vehicle's driving condition data.

[0007] Furthermore, the specific method for filtering the first category based on the slope difference of the vehicle's driving condition data includes: When the slope difference is greater than the first preset threshold, the type of vehicle driving status data corresponding to the vehicle driving status data at the target time point is the first type.

[0008] Furthermore, the specific method for dividing each row of data in the two-dimensional driving data matrix into multiple time periods includes: The first category, whose normalized retention value is greater than or equal to the first retention threshold, is used as the dividing point. Based on the dividing point, each row of data in the two-dimensional driving data matrix is ​​divided into multiple time periods.

[0009] Furthermore, the specific methods for obtaining the voting values ​​within a preset window scale during the time period of the target time point and determining the weight value of the benchmark voting value within the preset window scale include: The ratio of the number of time points within the preset window that are in the same time period as the target time point to the preset window size is recorded as the vote value of the target time point under that preset window size. The mean cosine similarity of the driving data vectors corresponding to the driving status data of all two adjacent cars within the preset window at the target time point is used as the first weight value of the preset window scale at the target time point, and the first weight value of the preset window scale is used as the weight value of the benchmark voting value under the preset window scale.

[0010] Furthermore, the specific methods for determining the mean first cosine similarity at the first time point and the mean second cosine similarity at the target time point are as follows: The data collection times corresponding to the driving status data of other cars in the time period of the target time point are respectively recorded as the first time point. The mean of the cosine similarity between the driving data vector corresponding to the first time point and the driving data vector corresponding to all data collection times in the time period of the target time point is recorded as the first cosine similarity mean of the first time point. The maximum value of the mean first cosine similarity of all data collection times within the time period of the target time point is recorded as the mean second cosine similarity of the target time point.

[0011] Furthermore, the specific method for obtaining the voting distance of the vehicle's driving status data based on the adjusted voting value, and marking important time periods based on the voting distance, includes: Using the largest third preset threshold adjustment vote value as the cluster center, density clustering is performed on the adjustment vote values ​​of all data collection times in the two-dimensional driving data matrix to obtain the third preset threshold clusters; The Euclidean distance between the adjusted voting value at the data collection time contained in the cluster and the cluster center is denoted as the voting distance of all car driving status data at the data collection time corresponding to the adjusted voting value; The time period in which the normalized mean of the voting distance of all included vehicle driving status data is greater than the second retention threshold is defined as an important time period.

[0012] Furthermore, the specific method for obtaining the feature weights of the first feature vector based on the influence weight values ​​and reconstructed influence weight values ​​of driving condition data of the same type of vehicles includes: in, For the first The feature weights of each feature vector; The number of types of data representing the vehicle's driving status; Indicates the first The influence weight values ​​of driving status data for each type of vehicle; Indicates the removal of the first The th eigenvector after the th eigenvector The reconstruction of driving status data for each type of vehicle affects the weight value; This represents the first adjustment coefficient.

[0013] Furthermore, the specific methods for determining the data to be retained and stored based on feature weights to achieve multi-dimensional intelligent data management for regional security management software include: The left singular matrix, right singular matrix, and diagonal matrix of the feature vectors whose normalized feature weights are greater than or equal to the third retention threshold are stored to realize multi-dimensional data intelligent management for regional security management software.

[0014] The beneficial effects of this invention are: This invention establishes a two-dimensional driving data matrix based on collected vehicle driving status data, performs SVD decomposition on the two-dimensional driving data matrix to obtain feature vectors, and stores the filtered feature vectors. This solves the problem of low efficiency in multi-dimensional data intelligent management caused by excessive redundant data, enabling efficient storage of multi-dimensional data used by regional safety management software. Specifically, it obtains a first category of retention degree based on the characteristic that the more drastic the change in sensor data values, the more frequent the system warnings, and the shorter the time interval between system warnings, the more effective information for regional safety warnings the sensor data values ​​contain. Based on the retention degree, the driving status data of the same type of vehicle is divided into time periods. Secondly, considering the influence of driving status data of other vehicles within the target time period, adjustment voting values ​​are obtained. Cluster centers are selected based on these adjusted voting values ​​to determine the degree of driving data anomaly, thereby improving the accuracy of the assessment of driving data anomaly. Finally, influence weight values ​​are determined, and feature weights of each eigenvector of the left and right singular matrices are determined based on these influence weight values. The data to be retained and stored is determined based on these feature weights, realizing multi-dimensional intelligent data management for regional safety management software. This can retain data related to regional safety early warning while ensuring a high data compression rate, solving the problem of low efficiency in multi-dimensional intelligent data management caused by excessive redundant data. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic flowchart of a multi-dimensional data intelligent management method for regional security management software provided in one embodiment of the present invention; Figure 2 Flowchart for adjusting the voting value acquisition process. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 The diagram illustrates a flowchart of a multi-dimensional data intelligent management method for regional security management software provided in an embodiment of the present invention. The method includes the following steps: Step S001: Collect vehicle driving status data, establish a two-dimensional driving data matrix, obtain the left singular matrix, right singular matrix, diagonal matrix, and eigenvectors of the left and right singular matrices of the two-dimensional driving data matrix, and obtain the number of system warnings and the time of each system warning.

[0019] This embodiment uses multidimensional data from area safety management software in the field of vehicle operation as an example for analysis.

[0020] Speed ​​sensors, acceleration sensors, and motor speed sensors are installed in areas requiring regional safety management. These sensors collect data on vehicle driving conditions within the area, including vehicle speed, acceleration, and motor speed. However, the types of vehicle driving condition data are not limited to vehicle speed, acceleration, and motor speed; implementers can set and select the types of data as needed. The data acquisition time and time interval for the speed, acceleration, and motor speed sensors are identical, with an empirical value of 3 seconds. A group of 600 data points of the same type from each vehicle is considered a single data point.

[0021] A corresponding two-dimensional driving data matrix is ​​established based on the driving status data of each group of cars. The driving status data of cars of the same type are arranged in the same row of the two-dimensional driving data matrix. The driving status data of cars of the same type are arranged from left to right in chronological order. The two-dimensional driving data matrix has three rows. From the first row to the third row, from top to bottom, they correspond to the car speed, the car acceleration, and the motor speed, respectively.

[0022] Because the multidimensional data used by the regional safety management software is massive, it needs to be compressed and stored in a database to facilitate subsequent data analysis and use, enabling safety warnings for vehicles within the area. However, the sensor data acquired in regional safety management is not always informative; much of it is redundant. Therefore, redundant data within each two-dimensional driving data matrix needs to be removed to improve data compression efficiency while retaining data relevant to regional safety warnings, thereby enhancing the intelligent management efficiency of multidimensional data.

[0023] The SVD decomposition algorithm is applied to the two-dimensional driving data matrix to obtain the eigenvectors of the left singular matrix, right singular matrix, diagonal matrix, and the left and right singular matrices. SVD decomposition of the two-dimensional driving data matrix is ​​a well-known technique and will not be elaborated further.

[0024] Extract the number of system alerts and the time of each system alert from the regional security early warning system in the regional security management software.

[0025] Thus, the two-dimensional driving data matrix, as well as the left singular matrix, right singular matrix, diagonal matrix, and eigenvectors of the left and right singular matrices of the two-dimensional driving data matrix, are obtained, along with the number of system warnings and the time of each system warning.

[0026] Step S002: Obtain driving data vectors based on the two-dimensional driving data matrix, determine the target time point based on the driving data vectors, obtain the slope difference of the vehicle's driving status data, filter the first category based on the slope difference of the vehicle's driving status data, obtain the retention degree of the first category, and divide each row of data in the two-dimensional driving data matrix into multiple time periods.

[0027] In order to improve compression efficiency while retaining vehicle driving status data containing regional safety warning information, it is necessary to filter the eigenvectors of the left and right singular matrices and remove eigenvectors with the same redundant information.

[0028] Each column in the two-dimensional driving data matrix is ​​treated as a driving data vector, and the data contained in the driving data vector corresponds to the driving status data of the car at the same data collection time.

[0029] Each driving data vector is taken as the driving data vector to be analyzed, and the data acquisition time corresponding to the data vector to be analyzed is recorded as the target time point.

[0030] Starting from the target time point, acquire vehicle driving status data for two data collection times after the target time point. Record the acquired vehicle driving status data as the adjacent driving status data of the target time point. Obtain the slope of the driving status data at the target time point and the two data collection times after the target time point. The slope of the driving status data is the absolute value of the difference between the driving status data at the next time point and the driving status data at the current time point. The average of the absolute values ​​of the differences between the driving status data slopes of the vehicle and each adjacent driving status data is recorded as the slope difference of the vehicle driving status data at the target time point.

[0031] When the slope difference of the vehicle's driving status data exceeds a first preset threshold, the type of vehicle's driving status data corresponding to the vehicle's driving status data at the target time point is considered to be the first type. The empirical value of the first preset threshold is 0.58.

[0032] Obtain the retention rate of the first type.

[0033] in, Indicates the first type The degree of preservation; Indicates the first type The number of the first category identified at the corresponding target time point; This represents the first adjustment coefficient, with an empirical value of 1; Indicates the first type In the two-dimensional driving data matrix, the first type Corresponding time and before, first type The number of times the system warnings appear for the corresponding types of vehicle driving status data; Indicates the first type In the two-dimensional driving data matrix, the first type The total number of system warnings that occurred at or before the corresponding time; Indicates the first type In the two-dimensional driving data matrix, it is related to the first type The time interval between the two system warnings that are closest in time; As the first category The corresponding sensor data values ​​are the same as those of the first type. The absolute value of the difference between the sensor data values ​​corresponding to the previous first category.

[0034] The larger the absolute value of the difference between the sensor data value corresponding to the first category and the previous sensor data value corresponding to the first category, the more drastic the change in the sensor data value corresponding to the first category. The more information the sensor data value corresponding to the first category contains that is effective for area safety warnings, and the greater the retention rate of the first category. In this case, the necessity of retaining the vehicle's driving status data at the time of collection for the first category is greater. Conversely, the more system warnings that occur before the time of collection for the first category, and the shorter the time interval between these warnings, the greater the likelihood that the sensor data value corresponding to the first category will be abnormal. The more information the sensor data value corresponding to the first category contains that is effective for area safety warnings, and the greater the retention rate of the first category. In this case, the necessity of retaining the vehicle's driving status data at the time of collection for the first category is greater.

[0035] The retention rates of all categories 1 are linearly normalized to obtain normalized values. When the normalized value of the retention rate of a category 1 is greater than or equal to the first retention rate threshold, the category 1 is marked as "retained," meaning that the vehicle driving status data at the data acquisition time corresponding to the "retained category" needs to be retained. The empirical value for the first retention rate threshold is 0.58.

[0036] Based on retaining the first category, the driving status data of the same type of vehicles are divided into multiple time periods, and each row of data in the two-dimensional driving data matrix is ​​divided into multiple time periods.

[0037] At this point, the driving data of the same type of vehicle is divided into time periods.

[0038] Step S003: Obtain the voting value within a preset window scale within the time period of the target time point; determine the weight value of the benchmark voting value within the preset window scale; perform a weighted average of the benchmark voting values ​​based on the weight value of the benchmark voting values ​​to obtain the benchmark voting value of the target time point; determine the mean of the first cosine similarity of the first time point; determine the mean of the second cosine similarity of the target time point; obtain the adjusted voting value based on the benchmark voting value of the target time point, the mean of the first cosine similarity of the first time point, and the mean of the second cosine similarity of the first time point; obtain the voting distance of the vehicle driving status data based on the adjusted voting value; and mark important time periods based on the voting distance.

[0039] Obtain the time period in which the target time point falls. Within the time period in which the target time point falls, obtain the baseline voting values ​​for all target time points under preset window scales. There are a total of two preset window scales, each with an empirical value of 5, 8, 11, 14, and 17.

[0040] Within each preset window size, the number of time points within the same time period as the target time point is counted, and the ratio of the counted number to the preset window size is recorded as the voting value of the target time point under that preset window size. The weighted average of the voting values ​​under the second preset threshold preset window sizes of the target time point is recorded as the baseline voting value of the target time point.

[0041] The weight value of the weighted average is determined as follows: the cosine similarity of the driving data vectors corresponding to the driving status data of every two adjacent cars within the preset window of the target time point is obtained, the mean of the cosine similarity of all adjacent driving data vectors within the preset window of the target time point is used as the first weight value of the preset window scale of the target time point, and the first weight value of the preset window scale is used as the weight value of the weighted average of the preset window scale.

[0042] Then, consider the impact of driving data of other vehicles in the same time period as the target time point.

[0043] The data collection times corresponding to the driving status data of other vehicles in the time period of the target time point are respectively recorded as the first time point. The mean of the cosine similarity between the driving data vector corresponding to the first time point and the driving data vector corresponding to all data collection times in the time period of the target time point is obtained. The mean of the cosine similarity corresponding to the first time point is recorded as the first cosine similarity mean of the first time point.

[0044] The maximum value of the mean first cosine similarity of all data collection times within the time period of the target time point is recorded as the mean second cosine similarity of the target time point.

[0045] The adjusted voting value is obtained based on the baseline voting value at the target time point, the mean of the first cosine similarity and the mean of the second cosine similarity at the first time point.

[0046] in, This indicates the adjusted vote value at the target time point; The mean first cosine similarity at the target time point; The mean second cosine similarity at the target time point; This represents the baseline vote value at the target time point.

[0047] When the mean first cosine similarity of the target time point is closer to the mean second cosine similarity of the target time point, and the base vote value of the target time point is larger, the adjustment vote value of the target time point is larger, the more information the vehicle driving status data corresponding to the target time point contains that is needed for regional safety management, and the greater the necessity to retain it.

[0048] The flowchart for adjusting the voting value is as follows: Figure 2 As shown.

[0049] The largest third preset threshold of adjusted votes is selected as the cluster center. The K-Means clustering algorithm is used to cluster the adjusted votes at all data collection times in the two-dimensional driving data matrix, obtaining the third preset threshold of clusters. The empirical value of the third preset threshold is 4.

[0050] Obtain the Euclidean distance between the adjusted voting value and the cluster center at the data collection time in each cluster. Record the Euclidean distance between the adjusted voting value and the cluster center as the voting distance of all car driving status data at the data collection time corresponding to the adjusted voting value.

[0051] The average voting distance of the vehicle driving status data within a time period is recorded as the degree of anomaly in the driving data for that time period. The greater the degree of anomaly in the driving data for a time period, the more information the driving status data contains for regional safety management, and the greater the necessity to retain it.

[0052] The time period in which the normalized value of the abnormality of the driving data is greater than the second retention threshold is defined as an important time period. The empirical value of the second retention threshold is 0.68.

[0053] This completes the acquisition of the crucial time period.

[0054] Step S004: Obtain the influence weight value of the driving status data of each type of vehicle, determine the first feature vector, obtain the reconstructed two-dimensional driving data matrix of the first feature vector, obtain the reconstruction influence weight value based on the reconstructed two-dimensional driving data matrix, obtain the feature weight of the first feature vector based on the influence weight value of the driving status data of the same type of vehicle and the reconstruction influence weight value, determine the data to be retained and stored based on the feature weight, and realize multi-dimensional intelligent data management for regional safety management software.

[0055] Obtain the influence weight value of driving condition data for each type of vehicle.

[0056] in, Indicates the first The influence weight values ​​of driving status data for each type of vehicle; This represents the first adjustment coefficient, with an empirical value of 1; In the two-dimensional driving data matrix, the first... The number of times the system warnings appear based on the driving status data of each type of vehicle; This indicates the number of system warnings that occur for all driving condition data contained in the two-dimensional driving data matrix; In the two-dimensional driving data matrix, the first... Among the system warnings that appear based on the driving status data of various types of vehicles, the time interval between the two closest adjacent system warnings; In the two-dimensional driving data matrix, the first... The number of important time periods included in the driving status data for each type of vehicle; Indicates the first Normalized values ​​of the degree of anomaly in driving data for key time periods; Indicates the first The variance of driving status data for all vehicles included within a key time period.

[0057] The greater the differences in driving status data for the same type of vehicle, and the greater the degree of anomaly in the driving data during important time periods for the same type of vehicle, the greater the influence weight of the driving status data for that type of vehicle. In other words, the more effective information for regional safety warnings the driving status data for that type of vehicle contains, the more it should be retained. Similarly, the more frequent the system warnings for the driving status data of the same type of vehicle, and the shorter the time interval between the two closest adjacent system warnings, the greater the influence weight of the driving status data for that type of vehicle. This also means that the more effective information for regional safety warnings the driving status data for that type of vehicle contains, the more it should be retained.

[0058] Each eigenvector of the left singular matrix and the right singular matrix is ​​taken as the first eigenvector.

[0059] After removing the left and right singular matrices corresponding to the first eigenvector, SVD reconstruction is performed on the remaining left and right singular matrices to obtain the reconstructed two-dimensional driving data matrix. The reconstructed two-dimensional driving data matrix is ​​denoted as the reconstructed two-dimensional driving data matrix of the first eigenvector.

[0060] Following the method described above for analyzing the two-dimensional driving data matrix to obtain the influence weight values ​​of driving status data for the same type of vehicle, the reconstruction influence weight values ​​of driving status data for the same type of vehicle are obtained based on the reconstructed two-dimensional driving data matrix.

[0061] The feature weights of the first feature vector are obtained by using the influence weights of driving condition data of the same type of vehicle and the reconstructed influence weights.

[0062] in, For the first The feature weights of each feature vector; The number of types of vehicle driving status data is 3 in this embodiment; Indicates the first The influence weight values ​​of driving status data for each type of vehicle; Indicates the removal of the first The th eigenvector after the th eigenvector The reconstruction of driving status data for each type of vehicle affects the weight value; This represents the first adjustment coefficient, with an empirical value of 1.

[0063] The more information the feature vector contains that is effective for regional safety warnings, the greater the difference between the influence weight value of the driving status data of the same type of vehicle and the reconstruction influence weight value, and the greater the feature weight of removing the feature vector.

[0064] Feature vectors with normalized feature weights greater than or equal to the third retention threshold are retained, while feature vectors with normalized feature weights less than the third retention threshold are removed. The left singular matrix, right singular matrix, and diagonal matrix of the retained feature vectors are stored. The empirical value of the third retention threshold is 0.38.

[0065] This completes the multi-dimensional intelligent management of data used in regional security management software.

[0066] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-dimensional data intelligent management method for regional security management software, characterized in that, The method includes the following steps: Collect vehicle driving status data, establish a two-dimensional driving data matrix, obtain the left singular matrix, right singular matrix, diagonal matrix, and eigenvectors of the left and right singular matrices of the two-dimensional driving data matrix, and obtain the number of system warnings and the time of each system warning; The driving data vector is obtained from the two-dimensional driving data matrix. The target time point is determined based on the driving data vector. The slope difference of the vehicle's driving condition data is obtained. The first category is filtered based on the slope difference of the vehicle's driving condition data. The retention degree of the first category is obtained. Each row of data in the two-dimensional driving data matrix is ​​divided into multiple time periods. Obtain the voting value within a preset window scale within the time period of the target time point; determine the weight value of the baseline voting value within the preset window scale; perform a weighted average of the baseline voting values ​​based on the weight values ​​of the baseline voting values ​​to obtain the baseline voting value of the target time point; determine the mean of the first cosine similarity at the first time point; determine the mean of the second cosine similarity at the target time point; obtain the adjusted voting value based on the baseline voting value of the target time point, the mean of the first cosine similarity at the first time point, and the mean of the second cosine similarity; obtain the voting distance of the vehicle driving status data based on the adjusted voting value; and mark important time periods based on the voting distance. The system obtains the influence weight value of driving status data for each type of vehicle, determines the first feature vector, obtains the reconstructed two-dimensional driving data matrix of the first feature vector, obtains the reconstruction influence weight value based on the reconstructed two-dimensional driving data matrix, obtains the feature weight of the first feature vector based on the influence weight value of driving status data for the same type of vehicle and the reconstruction influence weight value, and determines the data to be retained and stored based on the feature weight, thereby realizing multi-dimensional intelligent data management for regional safety management software.

2. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific methods for obtaining driving data vectors based on a two-dimensional driving data matrix and determining target time points based on driving data vectors are as follows: Each column in the two-dimensional driving data matrix is ​​treated as a driving data vector. Each driving data vector is taken as the driving data vector to be analyzed, and the data acquisition time corresponding to the data vector to be analyzed is recorded as the target time point.

3. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific method for obtaining the slope difference of the vehicle's driving status data is as follows: The vehicle's driving status data at two data collection times after the target time point will be recorded as the adjacent driving status data at the target time point; The absolute value of the difference between the driving condition data at the next moment and the current moment is recorded as the driving condition data slope at the current moment. The driving condition data slopes at the target time point and the two data collection moments after the target time point are obtained. The average of the absolute values ​​of the differences in the slopes of the driving condition data at the target time point and at two subsequent data collection times is recorded as the slope difference of the vehicle's driving condition data.

4. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific method for filtering the first category based on the slope difference of the vehicle's driving condition data is as follows: When the slope difference is greater than the first preset threshold, the type of vehicle driving status data corresponding to the vehicle driving status data at the target time point is the first type.

5. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific method for dividing each row of data in the two-dimensional driving data matrix into multiple time periods includes: The first category, whose normalized retention value is greater than or equal to the first retention threshold, is used as the dividing point. Based on the dividing point, each row of data in the two-dimensional driving data matrix is ​​divided into multiple time periods.

6. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific methods for obtaining the voting value within a preset window scale during the time period of the target time point and determining the weight value of the benchmark voting value within the preset window scale are as follows: The ratio of the number of time points within the preset window that are in the same time period as the target time point to the preset window size is recorded as the vote value of the target time point under that preset window size. The mean cosine similarity of the driving data vectors corresponding to the driving status data of all two adjacent cars within the preset window at the target time point is used as the first weight value of the preset window scale at the target time point, and the first weight value of the preset window scale is used as the weight value of the benchmark voting value under the preset window scale.

7. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific methods for determining the mean first cosine similarity at the first time point and the mean second cosine similarity at the target time point are as follows: The data collection times corresponding to the driving status data of other cars in the time period of the target time point are respectively recorded as the first time point. The mean of the cosine similarity between the driving data vector corresponding to the first time point and the driving data vector corresponding to all data collection times in the time period of the target time point is recorded as the first cosine similarity mean of the first time point. The maximum value of the mean first cosine similarity of all data collection times within the time period of the target time point is recorded as the mean second cosine similarity of the target time point.

8. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The method for determining the voting distance based on the adjusted voting value to obtain the vehicle's driving status data and marking important time periods based on the voting distance includes the following specific methods: Using the largest third preset threshold adjustment vote value as the cluster center, density clustering is performed on the adjustment vote values ​​of all data collection times in the two-dimensional driving data matrix to obtain the third preset threshold clusters; The Euclidean distance between the adjusted voting value at the data collection time contained in the cluster and the cluster center is denoted as the voting distance of all car driving status data at the data collection time corresponding to the adjusted voting value; The time period in which the normalized mean of the voting distance of all included vehicle driving status data is greater than the second retention threshold is defined as an important time period.

9. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific method for obtaining the feature weights of the first feature vector based on the influence weight values ​​and reconstructed influence weight values ​​of driving status data of the same type of vehicles includes: in, For the first The feature weights of each feature vector; The number of types of data representing the vehicle's driving status; Indicates the first The influence weight values ​​of driving status data for each type of vehicle; Indicates the removal of the first The th eigenvector after the th eigenvector The reconstruction of driving status data for each type of vehicle affects the weight value; This represents the first adjustment coefficient.

10. The multi-dimensional data intelligent management method for regional security management software according to claim 1, characterized in that, The specific methods for determining the data to be retained and stored based on feature weights to achieve multi-dimensional intelligent data management for regional security management software include: The left singular matrix, right singular matrix, and diagonal matrix of the feature vectors whose normalized feature weights are greater than or equal to the third retention threshold are stored to realize multi-dimensional data intelligent management for regional security management software.

Images