Parking behavior analysis method and system based on data mining and storage medium

Abstract

The application relates to the technical field of data processing, and discloses a parking behavior analysis method and system based on data mining and a storage medium. The method comprises the following steps: collecting data through a parking sensor network to construct a parking behavior basic data structure, a local interest point database and a global network configuration database; adopting a local-global feature interaction matrix algorithm to perform fusion processing to obtain a fusion feature matrix and a dynamic weight vector; performing clustering through a fusion density clustering algorithm to obtain a fusion clustering result matrix; and performing pattern recognition based on double-layer space-time mode mining to obtain a parking behavior space-time fusion mode library and a demand prediction result. The application solves the technical problem that an existing parking behavior analysis method based on data mining cannot effectively fuse local interest point features and global network configuration features, leading to a lack of significant correlation between analysis results and actual parking use modes.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a parking behavior analysis method, system and storage medium based on data mining. Background Technology

[0002] In existing technologies, parking behavior analysis methods mainly employ single-dimensional data mining techniques. For example, road network analysis methods based on spatial grammar evaluate the spatial configuration characteristics of parking areas by calculating indicators such as road integration and selectivity; parking hotspot identification methods based on the DBSCAN clustering algorithm identify spatial distribution patterns of parking space usage intensity through density clustering; and parking behavior analysis methods based on social media data mining identify parking preferences and sentiment tendencies by mining users' online behavior data. These existing technologies possess certain analytical capabilities within their respective fields, revealing some patterns and characteristics of parking behavior from different perspectives.

[0003] Existing methods often employ a single analytical dimension. Spatial grammar methods over-rely on global network configuration features while neglecting the influence of local points of interest (POIs), clustering analysis methods lack consideration for temporal dynamics, and social media analysis methods struggle to effectively correlate with actual physical parking data. Secondly, existing technologies lack effective feature fusion mechanisms, failing to balance the interaction between local factors (such as POI density and public transport accessibility) and global factors (such as road network integration and connectivity), leading to significant discrepancies between analytical results and actual parking usage patterns. Furthermore, existing methods suffer from insufficient depth in spatiotemporal pattern mining, making it difficult to simultaneously capture both short-term fluctuations and long-term trends in parking behavior. Summary of the Invention

[0004] This application provides a data mining-based parking behavior analysis method, system, and storage medium to address the technical problem that existing data mining-based parking behavior analysis methods cannot effectively integrate local interest point features and global network configuration features, resulting in a lack of significant correlation between the analysis results and actual parking usage patterns.

[0005] Firstly, this application provides a parking behavior analysis method based on data mining. The method includes: collecting and processing data from paid parking areas using a parking sensor network to obtain a basic parking behavior data structure, and constructing a local point of interest (POI) database and a global network configuration database; fusing the parking behavior data structure, the POI database, and the global network configuration database using a local-global feature interaction matrix algorithm to obtain a fused feature matrix and a dynamic weight vector; clustering the fused feature matrix and the dynamic weight vector using a fused density clustering algorithm to obtain a fused clustering result matrix; and performing pattern recognition processing based on the fused clustering result matrix using two-layer spatiotemporal pattern mining to obtain a parking behavior spatiotemporal fusion pattern library and demand prediction results.

[0006] Secondly, this application provides a parking behavior analysis system based on data mining, the parking behavior analysis system based on data mining includes:

[0007] The data acquisition module is used to collect and process data from the paid parking area through the parking sensor network, obtain the basic data structure of parking behavior, and build a local point of interest database and a global network configuration database.

[0008] The fusion module is used to perform fusion processing on the parking behavior basic data structure, local interest point database and global network configuration database through local-global feature interaction matrix algorithm to obtain fused feature matrix and dynamic weight vector;

[0009] The clustering module is used to cluster the fused feature matrix and dynamic weight vector using a fused density clustering algorithm to obtain a fused clustering result matrix.

[0010] The identification module is used to perform pattern recognition processing based on the fusion clustering result matrix through two-layer spatiotemporal pattern mining to obtain a spatiotemporal fusion pattern library of parking behavior and demand prediction results.

[0011] Thirdly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the aforementioned data mining-based parking behavior analysis method.

[0012] The technical solution provided in this application constructs a basic data structure for parking behavior, a local point of interest (POI) database, and a global network configuration database through data collection and processing of paid parking areas using a parking sensor network. This lays a multi-dimensional data foundation for subsequent feature fusion analysis, solving the problems of single data sources and incomplete information dimensions in existing technologies. It enables parking behavior analysis to simultaneously consider multiple influencing factors such as physical parking status, POI distribution, and road network configuration. The fused feature matrix and dynamic weight vector obtained through the local-global feature interaction matrix algorithm innovatively solve the core technical challenge of effectively fusing local and global features in existing technologies. This algorithm quantifies the interaction between local POI density features and global network integration features, and adopts a dynamic weight allocation mechanism to adaptively adjust the influence weights of local and global factors according to the feature distribution of different parking spaces. This significantly improves the accuracy and comprehensiveness of feature representation, allowing the fused feature matrix to more realistically reflect the comprehensive impact of parking spaces in complex urban environments. The fusion density clustering algorithm, by processing the fusion feature matrix and dynamic weight vector to obtain the fusion clustering result matrix, effectively solves the technical limitation of traditional clustering methods that cannot simultaneously identify local and global driving parking patterns. This algorithm dynamically adjusts the density parameters based on the local-global balance factor, enabling it to accurately identify both small-scale fine clusters influenced by POIs and large-scale macro-clusters influenced by network configuration within the same parking area, significantly improving clustering accuracy compared to traditional methods. The two-layer spatiotemporal pattern mining, through pattern recognition processing of the fusion clustering result matrix to obtain a spatiotemporal fusion pattern library of parking behavior and demand prediction results, overcomes the technical bottleneck of insufficient spatiotemporal pattern analysis depth in existing technologies. This method extracts POI influence patterns at the local driving layer and network configuration patterns at the global driving layer through a hierarchical processing mechanism, and organically combines local short-term fluctuations with global long-term trends through a pattern fusion algorithm, constructing a comprehensive prediction model that reflects both micro-behavioral characteristics and grasps macro-development laws.

[0013] In the application of parking behavior analysis, the core contribution of the local-global feature interaction matrix algorithm lies in its ability to automatically identify and quantify the interaction strength between features at different spatial scales. Through matrix multiplication and interaction coefficient training, this algorithm enables the parking behavior analysis system to accurately determine whether a parking space is more influenced by the local attraction of surrounding commercial facilities or by the global convenience of road network accessibility, thus providing a scientific basis for precise parking management strategies. The unique value of the fusion density clustering algorithm in urban parking management lies in its ability to simultaneously handle multi-scale parking clustering phenomena. This algorithm comprehensively considers parking duration similarity, location proximity, and POI influence similarity by fusing distance metrics, enabling the clustering results to identify both small-scale high-frequency use areas formed by commercial activity clusters and large-scale traffic distribution areas formed by transportation hubs. This provides a precise spatial division basis for differentiated parking space allocation and pricing strategies. The key role of the dual-layer spatiotemporal pattern mining algorithm in parking demand prediction lies in its ability to separate and reconstruct temporal patterns under different driving mechanisms. By performing time series decomposition of the POI impact on the local driving layer and time series analysis of network configuration on the global driving layer, the algorithm can accurately predict the changing trends of parking demand under specific spatiotemporal conditions. The local pattern captures short-term demand fluctuations affected by local factors such as commercial activities and public transportation, while the global pattern identifies long-term demand trends affected by global factors such as road network structure and urban development planning. The fusion prediction results of the two patterns provide data support for the dynamic allocation of parking resources and the optimized operation of intelligent guidance systems. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of an embodiment of the parking behavior analysis method based on data mining in this application.

[0016] Figure 2 This is a schematic diagram of one embodiment of the parking behavior analysis system based on data mining in this application. Detailed Implementation

[0017] This application provides a parking behavior analysis method, system, and storage medium based on data mining. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the parking behavior analysis method based on data mining in this application includes:

[0019] Step S101: Collect and process data from the paid parking area through the parking sensor network to obtain the basic data structure of parking behavior, and construct a local point of interest database and a global network configuration database.

[0020] Step S102: Based on the parking behavior basic data structure, the local interest point database and the global network configuration database, the local-global feature interaction matrix algorithm is used to perform fusion processing to obtain the fused feature matrix and dynamic weight vector;

[0021] Step S103: The fused feature matrix and dynamic weight vector are clustered using a fused density clustering algorithm to obtain a fused clustering result matrix;

[0022] Step S104: Based on the fusion clustering result matrix, perform pattern recognition processing through two-layer spatiotemporal pattern mining to obtain a spatiotemporal fusion pattern library of parking behavior and demand prediction results.

[0023] It is understood that the executing entity of this application can be a parking behavior analysis system based on data mining, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiment uses a server as an example for illustration.

[0024] Specifically, parking occupancy detection sensors record the occupancy status changes of each parking space at a frequency of 30 seconds, obtaining a raw parking dataset containing vehicle identification, location identification, start time, end time, and parking duration. This raw data is then processed by an outlier detection algorithm to remove faulty sensor data and fill in missing data using time series interpolation, forming a structured basic data structure for parking behavior. Simultaneously, based on the geographical coordinates of parking spaces, POI information for commercial facilities, public transportation stations, and office buildings within a 500-meter radius is collected and processed. The Euclidean distance from each parking space to the nearest point of interest is calculated, constructing a local point of interest database. Furthermore, spatial syntactic analysis is performed on the parking area according to the road network topology to calculate road connectivity index, integration degree, and selectivity parameters, constructing a global network configuration database.

[0025] The parking behavior data structure, including parking duration, location coordinates, and occupancy status, is processed to construct feature vectors. Combined with POI distance data from the local point of interest (POI) database and network integration data from the global network configuration database, a multi-dimensional feature vector containing five-dimensional feature values ​​is obtained. The local-global feature interaction matrix algorithm first extracts local POI density feature values ​​and global network integration feature values ​​from the multi-dimensional feature vector of parking behavior for feature separation, constructing local and global feature sub-vectors respectively, resulting in dimension-matched local and global feature matrices. Then, the local and global feature matrices are transposed and multiplied. Following matrix multiplication rules, the product of each local feature and each global feature is calculated to obtain the initial interaction relationship matrix. Based on historical parking data, the initial interaction relationship matrix is ​​trained with interaction coefficients. Regression analysis is used to calculate the influence strength of each feature combination on parking behavior, obtaining an interaction coefficient vector reflecting the influence weight of different feature combinations. The initial interaction relationship matrix and the interaction coefficient vector are then multiplied element-wise, with each element multiplied by its corresponding interaction coefficient value to obtain the feature interaction coefficient matrix. Next, the feature interaction coefficient matrix is ​​input into the sigmoid function for weight normalization. The ratio of local feature weights and global feature weights is dynamically adjusted according to the POI density and network integration of different parking spaces to obtain a dynamic weight vector with values ​​ranging from 0 to 1. The multidimensional feature vector of parking behavior is weighted and fused according to the dynamic weight vector. The local feature value is multiplied by the local weight and the global feature value is multiplied by the global weight, and the sum is calculated to obtain the fused feature matrix.

[0026] A fusion density clustering algorithm is used to calculate parking space density by inputting a fusion feature matrix and a dynamic weight vector. The density radius parameter of different parking areas is adjusted according to a local-global balance factor to obtain a density distribution map reflecting the degree of parking space clustering. Based on the density distribution map, core point identification is performed on parking spaces. Parking spaces with density values ​​exceeding a set threshold and containing the fewest parking spaces in their neighborhood are marked as core points, resulting in a core point set containing core parking space identifiers and density values. Based on the core point set, parking space similarity is calculated using a fusion distance metric function. The comprehensive similarity between parking spaces is calculated by comprehensively considering parking duration similarity, location proximity, and POI influence similarity, resulting in a similarity matrix reflecting the strength of parking space associations. The similarity matrix is ​​then input into a clustering expansion algorithm for parking area division. Starting from the core points, clustering is expanded to parking spaces with high similarity, grouping parking spaces with similar parking behavior patterns into the same cluster, resulting in a fusion clustering result matrix.

[0027] A two-layer spatiotemporal pattern mining method is employed to perform hierarchical clustering processing on the fused clustering result matrix based on local and global dominance. Clusters with high local dominance are classified into the local driving layer, and clusters with high global dominance are classified into the global driving layer, resulting in a hierarchical clustering matrix. Based on the hierarchical clustering matrix, the local driving layer undergoes POI impact time series decomposition processing to extract short-term fluctuation trends and periodic changes influenced by commercial facilities and public transportation, yielding a local spatiotemporal pattern set. Based on the hierarchical clustering matrix, the global driving layer undergoes network configuration time series analysis processing to identify long-term trend changes and macro-flow patterns influenced by road integration and connectivity, resulting in a global spatiotemporal pattern set. The local and global spatiotemporal pattern sets are input into a pattern fusion algorithm for interaction effect calculation. Through weighted fusion operations, a prediction model that comprehensively considers local short-term fluctuations and global long-term trends is constructed, resulting in a spatiotemporal fusion pattern library for parking behavior and demand prediction results.

[0028] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0029] The status of each parking space is monitored by a parking occupancy detection sensor. The changes in the occupancy status of the parking space are recorded at a 30-second sampling frequency to obtain the original parking dataset containing vehicle identification, location identification, start time, end time and parking duration.

[0030] The original parking dataset is input into an outlier detection algorithm for data cleaning, sensor fault data is removed, and time series interpolation is used to fill in the missing data, resulting in a structured parking behavior basic data structure.

[0031] Based on the geographic coordinates of parking spaces, POI information of commercial facilities, public transportation stations and office buildings within a 500-meter radius is collected and processed. The Euclidean distance from each parking space to the nearest point of interest is calculated to obtain a local point of interest database.

[0032] Based on the road network topology, spatial syntactic analysis is performed on the parking area to calculate the road connectivity index, integration degree, and selectivity parameters, thereby obtaining a global network configuration database that records network configuration characteristics.

[0033] Specifically, the parking occupancy detection sensor uses magnetic induction technology to detect the presence of metal vehicles above parking spaces. It performs a status scan every 30 seconds and records a timestamp. When a vehicle enters, it records the start time and vehicle identifier; when a vehicle leaves, it records the end time. The parking duration is calculated using the time difference. Simultaneously, the GPS coordinates of the parking space are combined to record the location identifier, forming a raw parking dataset containing five fields. The outlier detection algorithm uses statistical methods to identify abnormal data points. Specifically, it calculates the standard deviation and mean of the parking duration, marking data exceeding three times the normal range as outliers and removing them. Then, time series interpolation is used to complete the missing data. The interpolation algorithm performs linear interpolation based on the parking status change trend at adjacent time points, reorganizing the processed data into a structured parking behavior data structure.

[0034] POI information collection and processing utilizes a GIS geographic information system to acquire data on points of interest within a 500-meter radius of parking spaces. This includes the latitude and longitude coordinates, type identification, and importance rating of commercial facilities; the location information and number of public transportation stops; and attribute information such as the building area and number of resident companies in office buildings. Euclidean distance calculation employs a two-dimensional plane coordinate system. The straight-line distance between the parking space coordinates and each POI coordinate is calculated through square root operations. The nearest POI is selected as the primary influencing factor for that parking space, and the POI type, distance value, and influence weight are stored in the local point of interest database.

[0035] The spatial syntactic analysis process first constructs a road network topology graph of the parking area, using road intersections as nodes and road segments as edges to form a network graph structure. The road connectivity index is calculated by counting the number of connecting edges to each node. The integration parameter is calculated by analyzing the shortest path length from a given node to all other nodes in the network; a smaller value indicates stronger centrality of the node in the network. The selectivity parameter is calculated by counting the number of shortest paths passing through a given node, reflecting the strength of that node's mediating role in the network. These calculation results are then associated according to the road segments where parking spaces are located and stored in a global network configuration database.

[0036] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0037] The parking duration, location coordinates and occupancy status in the basic data structure of parking behavior are processed to construct feature vectors. Combined with the POI distance data in the local point of interest database and the network integration degree data in the global network configuration database, a multi-dimensional feature vector of parking behavior containing five-dimensional feature values ​​is obtained.

[0038] Based on the multidimensional feature vector of parking behavior, the interaction relationship is calculated and processed by the local-global feature interaction matrix algorithm. The local POI density feature and the global network integration feature are multiplied by the matrix and then multiplied by the interaction coefficient to obtain the feature interaction coefficient matrix that quantifies the mutual influence strength between local and global features.

[0039] The feature interaction coefficient matrix is ​​input into the sigmoid function for weight normalization. The ratio of local feature weights and global feature weights is dynamically adjusted according to the POI density and network integration of different parking spaces to obtain a dynamic weight vector with values ​​ranging from 0 to 1.

[0040] The multidimensional feature vector of parking behavior is weighted and fused based on the dynamic weight vector. The local feature value is multiplied by the local weight and the global feature value is multiplied by the global weight, and then summed to obtain the fused feature matrix containing the fused feature value.

[0041] Specifically, the feature vector construction process extracts parking duration from the basic parking behavior data structure and performs numerical standardization, converting the original duration data into standardized values ​​between 0 and 1. Simultaneously, it extracts the latitude and longitude coordinates of parking spaces and converts them into location vectors in a planar coordinate system. Occupancy status is converted into numerical features through binary encoding. POI distance data from the local point of interest database is transformed into distance feature values ​​through logarithmic transformation, and network integration degree data from the global network configuration database is directly used as global feature values. These five dimensions of feature values ​​are arranged sequentially to form a multi-dimensional feature vector for parking behavior.

[0042] The local-global feature interaction matrix algorithm first extracts local POI density features and global network integration degree features from the multi-dimensional feature vector of parking behavior, constructing two sub-vectors. Then, the local feature sub-vector is transposed and multiplied with the global feature sub-vector. Each local feature element is multiplied by each global feature element to obtain the initial interaction relationship value. The interaction coefficients are obtained through training on historical parking data, specifically using regression analysis to calculate the influence strength of different feature combinations on parking behavior. The regression coefficients are stored as interaction coefficient values ​​in the interaction coefficient vector. Each element of the initial interaction relationship matrix is ​​then multiplied element-wise with its corresponding interaction coefficient value to obtain the feature interaction coefficient matrix, which quantifies the strength of the mutual influence between local and global features.

[0043] Weight normalization involves inputting the feature interaction coefficient matrix into the sigmoid function for mathematical transformation. The sigmoid function uses exponential operations to map any real value to a probability value between 0 and 1. Specifically, each matrix element is substituted as an independent variable into the sigmoid function to obtain the normalized weight value. Dynamic weight adjustment allocates weights proportionally based on the POI density and network integration of different parking spaces. Parking spaces with higher POI density are assigned greater local feature weights, while those with higher network integration are assigned greater global feature weights, forming a dynamically adjusted weight vector.

[0044] The weighted fusion process multiplies the local feature values ​​of the multidimensional feature vector of parking behavior with their corresponding local weights, and simultaneously multiplies the global feature values ​​with their corresponding global weights. The two product results are then summed to obtain the fused feature value. After the fused feature values ​​for each parking space are calculated, they are arranged in order of parking space number to form a fused feature matrix. Each row in the matrix represents the fused feature of one parking space, and each column represents the fusion result of one feature dimension.

[0045] For example, the parking duration of a parking space is 3 hours. After standardization, the feature value is 0.6. After location coordinate transformation, the location feature values ​​are 0.8 and 0.3. The occupancy status is coded as 1. The POI distance is logarithmically transformed to obtain a feature value of 0.4. The network integration degree is 0.7, forming a five-dimensional feature vector. The local POI density feature 0.4 and the global network integration degree feature 0.7 are multiplied by a matrix to obtain an initial interaction value of 0.28. Multiplying this by the interaction coefficient 1.5 obtained from training with historical data, we get a feature interaction coefficient of 0.42. The interaction coefficient 0.42 is input into the sigmoid function to calculate the normalized weight 0.6. Based on the characteristic of this parking space having low POI density but high network integration, a local weight of 0.3 and a global weight of 0.7 are assigned. Finally, the local feature value 0.4 is multiplied by the local weight 0.3 to obtain 0.12, and the global feature value 0.7 is multiplied by the global weight 0.7 to obtain 0.49. The sum of these two values ​​gives a fused feature value of 0.61, which is stored in the fused feature matrix.

[0046] In one specific embodiment, the process of performing interaction relationship calculation based on the multi-dimensional feature vector of parking behavior using a local-global feature interaction matrix algorithm can specifically include the following steps:

[0047] Local POI density feature values ​​and global network integration feature values ​​are extracted from the multidimensional feature vector of parking behavior for feature separation processing. Local feature sub-vectors and global feature sub-vectors are constructed respectively to obtain dimension-matched local feature matrices and global feature matrices.

[0048] The local feature matrix and the global feature matrix are transposed and multiplied. The product value of each local feature and each global feature is calculated according to the matrix multiplication rules to obtain the initial interaction relation matrix.

[0049] Based on historical parking data, the interaction coefficients of the initial interaction relationship matrix are trained. The influence strength of each feature combination on parking behavior is calculated through regression analysis, and the interaction coefficient vector reflecting the influence weight of different feature combinations is obtained.

[0050] The initial interaction relationship matrix and the interaction coefficient vector are processed by element-wise multiplication. Each element in the matrix is ​​multiplied by the corresponding interaction coefficient value to obtain the feature interaction coefficient matrix that quantifies the strength of the mutual influence between local and global features.

[0051] Specifically, the feature separation process extracts local POI density feature values ​​and global network integration feature values ​​from the multi-dimensional feature vector of parking behavior according to preset feature index positions. The POI density feature values ​​reflect the density of points of interest around the parking space, while the network integration feature values ​​reflect the accessibility strength of the parking space in the road network. Local feature sub-vectors are constructed by traversing the POI density feature values ​​of all parking spaces and arranging them in parking space number order. Global feature sub-vectors are constructed by extracting the network integration feature values ​​of all parking spaces and arranging them in the same order. Dimension matching ensures that the local and global feature sub-vectors have the same dimension length. If the dimensions do not match, they are adjusted to the same dimension using zero-padding or truncation. Then, the local feature sub-vectors are expanded into row vectors to construct the local feature matrix, and the global feature sub-vectors are expanded into column vectors to construct the global feature matrix.

[0052] The transpose matrix product operation first transposes the local feature matrix, converting the original row vectors into column vectors. Then, following the matrix multiplication rules, the transposed local feature matrix is ​​multiplied by the global feature matrix. The matrix multiplication rule requires that the number of columns in the first matrix equals the number of rows in the second matrix. The product operation is performed by multiplying each element in each row of the first matrix by the corresponding element in each column of the second matrix, and then summing the results to obtain the corresponding element values ​​in the resulting matrix. The product value of each local feature and each global feature is calculated by directly multiplying the local feature value by its corresponding global feature value. All product values ​​are arranged according to their matrix positions to form an initial interaction matrix.

[0053] The interaction coefficient training process constructs a training dataset based on historical parking data. This historical data includes local and global feature values ​​for different parking spaces at different times, along with corresponding parking behavior outcomes such as parking duration and occupancy frequency. Regression analysis uses the least squares method to establish a mathematical model of the influence of combinations of local and global features on parking behavior. The influence strength of each feature combination on parking behavior is quantified by calculating regression coefficients. The influence strength calculation process uses local and global feature values ​​as independent variables and parking behavior outcomes as dependent variables, obtaining regression coefficients as interaction coefficient values. The interaction coefficient vector is constructed by arranging the regression coefficients corresponding to all feature combinations in matrix order, with each element in the vector corresponding to the interaction coefficient value at the same position in the initial interaction relationship matrix.

[0054] Element-wise multiplication involves multiplying the initial interaction relation matrix and the interaction coefficient vector element-wise at corresponding positions. Specifically, the element in the i-th row and j-th column of the initial interaction relation matrix is ​​multiplied by the corresponding interaction coefficient value in the interaction coefficient vector to obtain a new matrix element value. After the multiplication operation, a feature interaction coefficient matrix is ​​formed. Each element value in the matrix reflects the strength of the mutual influence between the corresponding local feature and the global feature combination; the larger the value, the stronger the mutual influence.

[0055] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0056] The fused feature matrix and dynamic weight vector are input into the fused density clustering algorithm to calculate the parking space density. The density radius parameter of different parking areas is adjusted according to the local-global balance factor to obtain a density distribution map that reflects the degree of parking space clustering.

[0057] Based on the density distribution map, the core point identification process of parking spaces is performed. Parking spaces with density values ​​exceeding a set threshold and containing the fewest number of parking spaces in their neighborhood are marked as core points, resulting in a set of core points containing core parking space identifiers and density values.

[0058] Based on the core point set, the similarity of parking spaces is calculated by fusing distance metrics. The comprehensive similarity between parking spaces is calculated by comprehensively considering the similarity of parking duration, location proximity and POI influence, and a similarity matrix reflecting the association strength of parking spaces is obtained.

[0059] The similarity matrix is ​​input into the clustering expansion algorithm for parking area division. Starting from the core point, the clustering is expanded to parking spaces with high similarity. Parking spaces with similar parking behavior patterns are merged into the same cluster, resulting in a fusion clustering result matrix containing cluster number, parking space list, local dominance, and global dominance.

[0060] Specifically, the fusion density clustering algorithm receives a fusion feature matrix and a dynamic weight vector as input data. First, it uses a density calculation function to statistically calculate the number of parking spaces in the neighborhood of each parking space. The density calculation searches for the number of other parking spaces within a set radius, centered on each parking space. A local-global balance factor dynamically adjusts the density radius parameter based on the ratio of the local feature intensity to the global feature intensity of each parking space in the fusion feature matrix. When the local features of a parking space dominate, the density radius is decreased to capture fine-grained local clustering patterns; when global features dominate, the density radius is increased to identify large-scale global clustering patterns. The density distribution map is generated by spatially mapping the density values ​​of all parking spaces according to their geographical coordinates. The color depth or numerical value of each location point in the map reflects the degree of parking space clustering at that location.

[0061] Core point identification is based on threshold judgment using density values ​​in the density distribution map. The set density threshold is calculated based on the average density and standard deviation of the overall parking space distribution. Parking spaces with density values ​​exceeding the threshold are included in the candidate core point set. The minimum number of parking spaces in the neighborhood check is performed by counting the number of other parking spaces within the neighborhood radius of the candidate core point. Only candidate points with the minimum number of parking spaces in their neighborhood are officially marked as core points. The core point set contains a unique identifier for each core parking space and its corresponding density value. The identifier is used for subsequent clustering expansion operations, and the density value is used to evaluate the importance of the core point.

[0062] The algorithm integrates distance metrics to comprehensively calculate the multi-dimensional similarity between parking spaces. Parking duration similarity is derived by normalizing the absolute value of the difference between the historical average parking durations of two parking spaces. Location proximity is calculated by inversely normalizing the Euclidean distance between the geographic coordinates of two parking spaces. POI influence similarity is calculated by comparing the similarity of the main POI types and distances around two parking spaces. The comprehensive similarity calculation involves weighted summation of the three similarity values ​​according to preset weights, which are dynamically adjusted based on the characteristic distribution of different parking areas. The similarity matrix is ​​constructed by calculating the comprehensive similarity values ​​between all pairs of parking spaces. The rows and columns of the matrix represent different parking spaces, and the matrix element values ​​represent the association strength between corresponding parking space pairs. The clustering expansion algorithm starts the expansion process with each core point in the core point set as a cluster seed. The expansion operation queries the similarity matrix to find other parking spaces with a similarity exceeding a set threshold to the current core point and adds these parking spaces to the same cluster. The expansion process employs a breadth-first search strategy. If a newly added parking space also meets the core point criteria, the expansion continues outwards until no new parking spaces meet the inclusion criteria. Cluster numbers are assigned sequentially according to the order in which the core points were discovered. The parking space list records all parking space identifiers contained in each cluster. Local dominance is calculated by statistically analyzing the mean of local feature values ​​of parking spaces within a cluster, while global dominance is calculated by calculating the mean of global feature values ​​of parking spaces within a cluster. These two dominance values ​​reflect the relative strength of the cluster's influence from local and global factors.

[0063] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0064] The fused clustering result matrix is ​​then processed into a hierarchical clustering layer based on local dominance and global dominance. Clusters with high local dominance are classified into the local driving layer, and clusters with high global dominance are classified into the global driving layer, resulting in a hierarchical clustering matrix containing two-layer classification labels.

[0065] Based on the hierarchical clustering matrix, the local driving layer is decomposed into POI impact time series to extract short-term fluctuation trends and periodic changes affected by commercial facilities and public transportation, resulting in a set of local spatiotemporal patterns that reflect the driving laws of POI.

[0066] Based on the hierarchical clustering matrix, the network configuration time series analysis of the global driving layer is performed to identify the long-term trend changes and macro-flow patterns affected by road integration and connectivity, and to obtain a set of global spatiotemporal patterns that reflect the influence of network configuration.

[0067] The local and global spatiotemporal pattern sets are input into the pattern fusion algorithm for interactive effect calculation. Through weighted fusion operation, a prediction model that comprehensively considers local short-term fluctuations and global long-term trends is constructed, resulting in a spatiotemporal fusion pattern library of parking behavior and demand prediction results containing spatiotemporal fusion patterns and parking demand prediction data.

[0068] Specifically, the hierarchical clustering process determines the hierarchy by comparing the local dominance and global dominance values ​​of each cluster in the fused clustering result matrix. When the local dominance value of a cluster is greater than the global dominance value, the cluster is marked as a local driving layer; when the global dominance value is greater than the local dominance value, the cluster is marked as a global driving layer. The two-level classification identifier is generated by adding a hierarchical identifier to the original cluster number. The local driving layer cluster identifier is "L-cluster number," and the global driving layer cluster identifier is "G-cluster number," forming a hierarchical clustering matrix containing hierarchical information.

[0069] The POI-based time series decomposition process extracts cluster data labeled as local driving layers from a hierarchical clustering matrix, obtaining historical parking occupancy time series data for each parking space in these clusters. Then, the time series data is classified and grouped according to the type of commercial facilities and public transportation stations surrounding the parking spaces. The time series decomposition uses an additive decomposition model to divide the original time series into three parts: trend component, periodic component, and random component. The trend component is calculated using the moving average method to reflect the long-term direction of change. The periodic component uses a periodic detection algorithm to identify repeating patterns such as daily and weekly cycles. The random component is the residual after subtracting the trend and periodic components from the original data. Short-term fluctuation trends are extracted by analyzing high-frequency change features in the periodic component, and periodic changes are identified by statistically analyzing repeating patterns at different time scales. These features are classified and stored in a local spatiotemporal pattern set according to POI type.

[0070] Network configuration time series analysis extracts cluster data labeled as the global driving layer from the hierarchical clustering matrix, obtains historical parking flow time series data for each parking space in these clusters, and performs correlation analysis by combining the integration degree and connectivity parameters of the roads where the parking spaces are located. Long-term trend changes are identified through linear regression analysis of the time series data; the regression slope reflects the long-term growth or decline trend of parking demand, and the regression intercept reflects the basic parking demand level. Macro-flow patterns are identified by analyzing parking flow transfer patterns between different road network nodes, calculating the flow correlation coefficient between parking spaces to quantify the mutual influence strength between nodes, and classifying and storing these patterns according to network configuration characteristics in a global spatiotemporal pattern set.

[0071] The pattern fusion algorithm receives local and global spatiotemporal pattern sets as input data to calculate interaction effects. These interaction effects are quantified by calculating the correlation coefficient between local and global pattern features. The weighted fusion operation multiplies local short-term fluctuation features by dynamically assigned local weights and global long-term trend features by their corresponding global weights. The two weighted results are then summed to obtain the fused prediction value. Weight allocation is dynamically adjusted based on the characteristics of different time periods and spatial regions; local weights are increased during weekdays to highlight the impact of Points of Interest (POIs), while global weights are increased during weekends to highlight the impact of network flow. The parking behavior spatiotemporal fusion pattern library is constructed by storing fusion pattern features under different spatiotemporal conditions. Demand prediction results are calculated by inputting the current spatiotemporal conditions into the fusion prediction model.

[0072] For example, commercial areas can be categorized into locally driven clusters centered around large shopping malls and globally driven clusters centered around major road intersections. Hierarchical processing labels the parking clusters around shopping malls as locally driven layers due to their high local dominance, and the parking clusters at road intersections as globally driven layers due to their high global dominance. Time-series decomposition of the POI impact of the locally driven layers reveals that the shopping mall cluster exhibits a clear shopping-oriented fluctuation pattern on weekday evenings and throughout the day on weekends, while the subway station cluster shows a commuter-oriented bi-peak pattern during weekday morning and evening rush hours. These patterns are further analyzed by time-series decomposition to extract corresponding periodic and trend components. Network configuration analysis of the globally driven layers reveals that parking demand in the major road intersection cluster is highly correlated with changes in overall urban traffic flow, and its long-term trend follows changes in road network capacity. Macro-flow patterns show a clear transfer relationship of parking demand between different intersections. The pattern fusion algorithm weights and merges local shopping and commuting cycle patterns with global traffic flow trend patterns. It increases the weight of local patterns during peak shopping periods and increases the weight of global patterns during traffic congestion periods. The final fusion prediction model can simultaneously capture short-term demand fluctuations driven by local business activities and long-term demand trends driven by the global transportation network.

[0073] In one specific embodiment, the process of performing POI influence time series decomposition processing on the local driving layer based on the hierarchical clustering matrix can specifically include the following steps:

[0074] The parking clusters with high local dominance values ​​are selected from the hierarchical clustering matrix for data extraction and processing to obtain the historical occupancy time series of parking spaces and POI distance information, thus obtaining a local driving parking time series dataset.

[0075] The local driving parking time series dataset is classified according to POI type, and the data groups are divided according to the impact of commercial facilities and public transportation stations, resulting in a time series data subset classified by POI type;

[0076] Frequency domain decomposition is performed on a subset of time-series data using Fourier transform to extract the spectral components of daily, weekly, and monthly periods, resulting in a frequency domain feature matrix containing periodic characteristics.

[0077] The temporal reconstruction process is performed based on the frequency domain feature matrix, and the patterns are classified according to the influence intensity of POIs to obtain a set of local spatiotemporal patterns containing the driving laws of POIs.

[0078] Specifically, the data extraction and processing identifies parking clusters with values ​​higher than a set threshold by scanning the local dominance field in the hierarchical clustering matrix. Then, based on the cluster number, it extracts the occupancy time series data of the corresponding parking spaces from the historical parking database. The time series data records the occupancy status changes of each parking space in hourly increments. POI distance information is extracted from the local point of interest database based on parking space identifiers, including the distance value from each parking space to the nearest commercial facility or public transportation stop and the POI type identifier. The time series data and POI information are then correlated and matched according to the parking space number to form a locally driven parking time series dataset.

[0079] The POI type classification process divides the localized parking time-series dataset into different data groups based on the type identifier in the POI distance information. The commercial facility impact group contains parking space data closest to shopping malls, restaurants, and entertainment venues, while the public transportation station impact group contains parking space data closest to subway stations and bus stops. The classification process determines the dominant influencing factors by comparing the distance values ​​of each parking space to different POI types, assigning the parking spaces to the nearest POI type group, forming time-series data subsets classified by POI type. Each data subset contains time-series data of multiple parking spaces under the influence of the same POI type.

[0080] Fourier transform processing converts each time series in the time-series data subset from the time domain to the frequency domain for periodic analysis. The transformation process identifies periodic components by calculating the amplitude and phase information of the time series at different frequencies. Daily periodic spectral components are extracted by detecting frequency components of a 24-hour repeating pattern, weekly periodic spectral components by detecting frequency components of a 7-day repeating pattern, and monthly periodic spectral components by detecting frequency components of a 30-day repeating pattern. Frequency domain decomposition breaks down the original time series into combinations of sine and cosine waves of different frequencies. The amplitude value of each frequency component reflects the intensity of the periodic pattern, and the phase value reflects the time shift of the periodic pattern. The frequency domain feature matrix is ​​constructed by arranging the spectral components of all parking spaces according to frequency and parking space number; the rows of the matrix represent different frequencies, and the columns represent different parking spaces.

[0081] Temporal reconstruction processing uses inverse Fourier transform to convert the main frequency components in the frequency domain feature matrix back to the time domain signal. The reconstruction process selects frequency components with larger amplitude values ​​for superposition calculation to reconstruct the time series. The influence intensity of POIs is determined by calculating the correlation between the reconstructed signal and the distance to the POI. Parking spaces closer to the POI have larger amplitudes of the corresponding periodic components in their reconstructed signals, indicating a stronger influence from that POI. Pattern classification categorizes the reconstructed time series patterns based on the influence intensity of different POI types. Commercial POI-driven patterns exhibit high-frequency usage characteristics on weekends and evenings, while traffic POI-driven patterns exhibit a bimodal characteristic during weekday morning and evening rush hours. These pattern features are stored in a local spatiotemporal pattern set.

[0082] The above describes the parking behavior analysis method based on data mining in the embodiments of this application. The following describes the parking behavior analysis system based on data mining in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the parking behavior analysis system based on data mining in this application includes:

[0083] The data acquisition module 201 is used to collect and process data from the paid parking area through the parking sensor network, obtain the basic data structure of parking behavior, and build a local point of interest database and a global network configuration database.

[0084] The fusion module 202 is used to perform fusion processing on the parking behavior basic data structure, local interest point database and global network configuration database through local-global feature interaction matrix algorithm to obtain fused feature matrix and dynamic weight vector;

[0085] Clustering module 203 is used to cluster the fused feature matrix and dynamic weight vector using a fused density clustering algorithm to obtain a fused clustering result matrix;

[0086] The identification module 204 is used to perform pattern recognition processing based on the fusion clustering result matrix through two-layer spatiotemporal pattern mining to obtain a spatiotemporal fusion pattern library of parking behavior and demand prediction results.

[0087] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the data mining-based parking behavior analysis method.

[0088] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data mining-based parking behavior analysis method, characterized by, The method comprises: The method comprises: According to the parking behavior basic data structure, the local interest point database and the global network configuration database, the fusion feature matrix and the dynamic weight vector are obtained by local-global feature interaction matrix algorithm, comprising: the parking duration, position coordinates and occupancy state in the parking behavior basic data structure are subjected to feature vector construction processing, the POI distance data in the local interest point database and the network integration degree data in the global network configuration database are combined, and a parking behavior multi-dimensional feature vector containing five-dimensional feature values is obtained; based on the parking behavior multi-dimensional feature vector, the local-global feature interaction matrix algorithm is used for interaction relationship calculation processing, the local POI density feature and the global network integration degree feature are subjected to matrix multiplication operation and multiplied by an interaction coefficient, and a feature interaction coefficient matrix quantifying the mutual influence strength of the local and global features is obtained; the feature interaction coefficient matrix is input into a sigmoid function for weight normalization processing, the POI density and the network integration degree of different parking spaces are used for dynamically adjusting the local feature weight and the global feature weight ratio, and a dynamic weight vector with a value range of 0 to 1 is obtained; according to the dynamic weight vector, the parking behavior multi-dimensional feature vector is subjected to weighted fusion processing, the local feature value is multiplied by the local weight, the global feature value is multiplied by the global weight, and summation operation is performed, and a fusion feature matrix containing fusion feature values is obtained; The fusion feature matrix and the dynamic weight vector are clustered by a fusion density clustering algorithm to obtain a fusion clustering result matrix, including: the fusion feature matrix and the dynamic weight vector are input into the fusion density clustering algorithm for parking space density calculation processing, the density radius parameters of different parking areas are adjusted according to the local-global balance factor to obtain a density distribution graph reflecting the parking space aggregation degree; based on the density distribution graph, core point recognition processing is performed on the parking space, parking spaces with density values exceeding a set threshold and containing the least number of parking spaces in the neighborhood are marked as core points to obtain a core point set containing core parking space identifiers and density values; the core point set is subjected to parking space similarity calculation processing by a fusion distance measurement function, and the comprehensive similarity between parking spaces is calculated by comprehensively considering the parking time length similarity, the location proximity and the POI influence similarity, to obtain a similarity matrix reflecting the parking space correlation strength; the similarity matrix is input into a clustering expansion algorithm for parking area division processing, and the clustering expansion is performed from the core point to the parking spaces with high similarity, and the parking spaces with similar parking behavior patterns are merged into the same cluster to obtain a fusion clustering result matrix containing cluster numbers, parking space lists, local dominant degrees and global dominant degrees, wherein the local dominant degree is calculated by statistically calculating the mean value of the local characteristic values of the parking spaces in the cluster, and the global dominant degree is calculated by calculating the mean value of the global characteristic values of the parking spaces in the cluster; Based on the fusion clustering result matrix, mode recognition processing is performed by double-layer spatio-temporal pattern mining to obtain a parking behavior spatio-temporal fusion pattern library and a demand prediction result, including: the fusion clustering result matrix is subjected to clustering layering processing according to the local dominant degree and the global dominant degree, the clustering with a high local dominant degree is divided into a local driving layer, and the clustering with a high global dominant degree is divided into a global driving layer to obtain a layered clustering matrix containing double-layer classification identifiers; the local driving layer is subjected to POI influence time series decomposition processing based on the layered clustering matrix, and the short-term fluctuation trend and the periodic change affected by commercial facilities and public transportation are extracted to obtain a local spatio-temporal pattern set reflecting the POI driving law; the global driving layer is subjected to network configuration time series analysis processing according to the layered clustering matrix, and the long-term trend change and the macro flow law affected by the road integration degree and the connectivity are identified to obtain a global spatio-temporal pattern set reflecting the network configuration influence; the local spatio-temporal pattern set and the global spatio-temporal pattern set are input into a pattern fusion algorithm for interactive effect calculation processing, a prediction model considering the local short-term fluctuation and the global long-term trend is constructed by weighted fusion operation to obtain a parking behavior spatio-temporal fusion pattern library and a demand prediction result containing spatio-temporal fusion laws and parking demand prediction data.

2. The data mining-based parking behavior analysis method according to claim 1, characterized in that, The local POI density feature and the global network integration degree feature are subjected to matrix multiplication operation and multiplied by an interaction coefficient based on the parking behavior multi-dimensional feature vector by a local-global feature interaction matrix algorithm to obtain a feature interaction coefficient matrix quantifying the mutual influence strength of the local and global features, including: Extracting local POI density feature values and global network integration degree feature values from the parking behavior multi-dimensional feature vector for feature separation processing, respectively constructing local feature sub-vectors and global feature sub-vectors, and obtaining dimensionally matched local feature matrices and global feature matrices; Performing transpose matrix multiplication operation processing on the local feature matrices and the global feature matrices, calculating the product value of each local feature and each global feature according to the matrix multiplication rule, and obtaining an initial interaction relationship matrix; Based on historical parking data, performing interaction coefficient training processing on the initial interaction relationship matrix, calculating the influence strength of each feature combination on parking behavior through regression analysis, and obtaining an interaction coefficient vector reflecting the influence weight of different feature combinations; Performing element-by-element multiplication operation processing on the initial interaction relationship matrix and the interaction coefficient vector, multiplying each element in the matrix by the corresponding interaction coefficient value, and obtaining a feature interaction coefficient matrix quantifying the mutual influence strength of local and global features.

3. The data mining based parking behavior analysis method of claim 1, wherein, Based on the hierarchical clustering matrix, performing POI influence time series decomposition processing on the local driving layer, extracting short-term fluctuation trends and periodic changes affected by commercial facilities and public transportation, and obtaining a local spatiotemporal pattern set reflecting the POI driving law, including: From the hierarchical clustering matrix, filtering parking clusters with high local dominant degree for data extraction processing, obtaining parking occupancy time series and POI distance information, and obtaining a local driving parking time series dataset; Classifying the local driving parking time series dataset according to POI types, dividing data groups according to the influence of commercial facilities and public transportation sites, and obtaining time series data subsets classified by POI types; Based on the time series data subsets, performing frequency domain decomposition processing through Fourier transform, extracting daily, weekly, and monthly frequency spectrum components, and obtaining a frequency domain feature matrix containing periodic characteristics; According to the frequency domain feature matrix, performing time domain reconstruction processing, classifying modes according to POI influence strength, and obtaining a local spatiotemporal pattern set containing POI driving law.

4. A data mining-based parking behavior analysis system, characterized by, A data mining-based parking behavior analysis system for implementing any one of claims 1-3, the data mining-based parking behavior analysis system comprising: The collection module is used for collecting and processing data of the paid parking area through the parking sensor network, obtaining a parking behavior basic data structure, constructing a local point of interest database and a global network configuration database, and includes: performing state monitoring and processing on each parking space through a parking occupancy detection sensor, recording parking space occupancy state changes at a 30-second collection frequency, and obtaining an original parking data set containing vehicle identification, location identification, start time, end time and parking duration; inputting the original parking data set into an outlier detection algorithm for data cleaning processing, eliminating sensor fault data and filling in missing data using a time series interpolation method, and obtaining a structured parking behavior basic data structure; collecting and processing POI information of commercial facilities, public transportation sites and office buildings within a 500-meter range based on parking space geographic coordinates, calculating the Euclidean distance from each parking space to the nearest point of interest, and obtaining a local point of interest database; performing spatial syntax analysis and processing on the parking area according to the road network topology structure, calculating road connectivity index, integration degree and selection degree parameters, and obtaining a global network configuration database recording network configuration characteristics; The fusion module is used for fusion processing according to the parking behavior basic data structure, the local point of interest database and the global network configuration database through a local-global feature interaction matrix algorithm, obtaining a fusion feature matrix and a dynamic weight vector, and includes: performing feature vector construction processing on the parking duration, location coordinates and occupancy state in the parking behavior basic data structure, combining the POI distance data in the local point of interest database and the network integration degree data in the global network configuration database, and obtaining a five-dimensional feature value containing a parking behavior multi-dimensional feature vector; performing interaction relationship calculation processing based on the parking behavior multi-dimensional feature vector through the local-global feature interaction matrix algorithm, performing matrix multiplication operation on the local POI density feature and the global network integration degree feature and multiplying by an interaction coefficient, and obtaining a feature interaction coefficient matrix quantifying the mutual influence strength of local and global features; inputting the feature interaction coefficient matrix into a sigmoid function for weight normalization processing, dynamically adjusting the local feature weight and the global feature weight proportion according to the POI density and the network integration degree of different parking spaces, and obtaining a dynamic weight vector with a value range of 0 to 1; performing weighted fusion processing on the parking behavior multi-dimensional feature vector according to the dynamic weight vector, performing summation operation on the local feature value multiplied by the local weight and the global feature value multiplied by the global weight, and obtaining a fusion feature matrix containing a fusion feature value; The clustering module is configured to perform clustering processing on the fusion feature matrix and the dynamic weight vector by using a fusion density clustering algorithm to obtain a fusion clustering result matrix, including: inputting the fusion feature matrix and the dynamic weight vector into the fusion density clustering algorithm to perform parking space density calculation processing, adjusting the density radius parameters of different parking areas according to a local-global balance factor to obtain a density distribution graph reflecting the parking space aggregation degree; performing core point identification processing on the parking spaces based on the density distribution graph, marking a parking space as a core point if the density value of the parking space exceeds a set threshold and the parking space contains the least number of parking spaces in the neighborhood, to obtain a core point set containing core parking space identifiers and density values; performing parking space similarity calculation processing on the core point set by using a fusion distance measurement function, and comprehensively considering the parking duration similarity, the location proximity and the POI influence similarity to calculate the comprehensive similarity between the parking spaces, to obtain a similarity matrix reflecting the parking space correlation strength; inputting the similarity matrix into a clustering expansion algorithm to perform parking area division processing, taking the core points as starting points to expand the clustering to parking spaces with high similarity, and merging parking spaces with similar parking behavior patterns into the same cluster to obtain a fusion clustering result matrix containing a cluster number, a parking space list, a local dominant degree and a global dominant degree, wherein the local dominant degree is calculated by statistically calculating the mean value of the local feature values of the parking spaces in the cluster, and the global dominant degree is calculated by calculating the mean value of the global feature values of the parking spaces in the cluster. The identification module is configured to perform pattern identification processing on the fusion clustering result matrix by using a double-layer spatio-temporal pattern mining to obtain a parking behavior spatio-temporal fusion pattern library and a demand prediction result, including: performing clustering and layering processing on the fusion clustering result matrix according to the local dominant degree and the global dominant degree, dividing the clusters with high local dominant degrees into a local driving layer, and dividing the clusters with high global dominant degrees into a global driving layer, to obtain a layered clustering matrix containing double-layer classification identifiers; performing POI influence time series decomposition processing on the local driving layer based on the layered clustering matrix, extracting short-term fluctuation trends and periodic changes affected by commercial facilities and public transportation, to obtain a local spatio-temporal pattern set reflecting the POI driving law; performing network configuration time series analysis processing on the global driving layer based on the layered clustering matrix, identifying long-term trend changes and macro flow rules affected by road integration degree and connectivity, to obtain a global spatio-temporal pattern set reflecting the network configuration influence; inputting the local spatio-temporal pattern set and the global spatio-temporal pattern set into a pattern fusion algorithm to perform interactive effect calculation processing, constructing a prediction model that comprehensively considers the local short-term fluctuation and the global long-term trend by using weighted fusion operation, to obtain a parking behavior spatio-temporal fusion pattern library and a demand prediction result containing spatio-temporal fusion rules and parking demand prediction data.

5. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by a processor, causes the processor to perform the data mining-based parking behavior analysis method according to any one of claims 1 to 3.

Images