A smart center space-time big data intelligent scheduling analysis method and system

Abstract

The application relates to the field of big data analysis and intelligent computing technology, in particular to a smart computing center space-time big data intelligent scheduling analysis method and system; the method comprises the following steps: short-term and long-term behavior characteristics of tourists are extracted through multi-source data normalization and semanticized trajectory mapping, a cross-scale coupling matrix is constructed, and a unified fractal feature vector is generated; on the basis, regional behavior entropy is calculated and drives adaptive adjustment of the partition, hot areas are dynamically identified in combination with tourist flow and entropy contribution rate; further, a behavior driving potential field is constructed by fusing regional fractal features, behavior entropy and historical flow sequences, multi-scale adaptive prediction and error feedback optimization are carried out; finally, a scheduling target function is constructed based on the prediction result, multi-dimensional candidate strategies are generated, and intelligent scheduling of scenic resources and optimization of tourist distribution are realized through comprehensive scoring and closed-loop optimization. The application improves the operation efficiency of the scenic area and the tourist experience.

Description

Technical Field

[0001] This invention relates to the field of big data analysis and intelligent computing technology, specifically to a method and system for intelligent scheduling and analysis of spatiotemporal big data in an intelligent computing center. Background Technology

[0002] With the continuous improvement of smart tourism and urban intelligent management, the operation and management of scenic spots and large public places are gradually developing towards digitalization and refinement.

[0003] Chinese invention patent application CN120259028A discloses a destination-wide tourism big data management platform based on data processing. The platform includes a perception aggregation grid module that collects tourism data and constructs a geographic grid network, sets spatial grid reconstruction rules, and automatically refines or merges grids. It also introduces a trajectory triggering mechanism that, when the frequency of behavior within a preset target area reaches a preset behavior frequency threshold, enables fine-grained modeling to generate a geographic perception structure. A behavior deconstruction and encoding module analyzes tourists' continuous behavior within the geographic perception structure and constructs a multi-dimensional contextual encoding vector. Finally, it reconstructs the complete behavior chain of tourists through a behavior jigsaw puzzle-style deconstruction method and an implicit behavior compensation method.

[0004] The widespread adoption of multi-source sensing devices, IoT technology, and positioning systems has provided massive amounts of spatiotemporal data support for tourist behavior analysis. Against this backdrop, how to efficiently integrate multi-source heterogeneous data, deeply mine multi-scale behavioral characteristics of tourists, and achieve dynamic optimization and scheduling of scenic area resources has become an important research direction in the field of smart management. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing a method and system for intelligent scheduling and analysis of spatiotemporal big data in intelligent computing centers.

[0006] The technical solution of this invention: A method for intelligent scheduling and analysis of spatiotemporal big data in a smart computing center, comprising the following specific implementation steps:

[0007] S1. Extract short-term and long-term behavioral features of tourists through multi-source data normalization and semantic trajectory mapping, construct a cross-scale coupling matrix and generate a unified fractal feature vector;

[0008] S2. By projecting fractal features onto the regional level, calculating regional behavioral entropy and driving adaptive adjustment of partitions, hotspot areas are dynamically identified by combining tourist flow and entropy contribution rate.

[0009] S3. Integrate regional fractal features, behavioral entropy and historical flow sequences to construct a behavior-driven potential field and perform multi-scale adaptive prediction and error feedback optimization.

[0010] S4. Based on the prediction results, construct the scheduling objective function, generate multi-dimensional candidate strategies, and perform intelligent scheduling of scenic resources and optimization of tourist distribution through comprehensive scoring and closed-loop optimization.

[0011] Preferably, multi-source data normalization and semantic trajectory mapping includes:

[0012] Collect multi-source heterogeneous data within the scenic area, including tourist movement trajectories, gate entry and exit records, ticket reservation times, surveillance video counts, crowd heat maps, weather and event data;

[0013] The data is cleaned and normalized, and a confidence weight is assigned to each data point.

[0014] The trajectory points are mapped to semantic units, the dwell time and behavior category statistics of each unit are calculated, and the semantic trajectory sequence is output.

[0015] Preferably, constructing a cross-scale coupling matrix and generating a unified fractal eigenvector includes:

[0016] An entropy-driven adaptive fractal scale selection method dynamically selects the effective scale set for each trajectory.

[0017] Statistical and information-theoretic features are extracted from trajectory segments at each scale, and segment feature vectors are output.

[0018] Construct an inter-scale coupling matrix to quantify the correlation of features at different time scales;

[0019] Multi-scale segment features and cross-scale coupled features are fused into a unified fractal feature vector.

[0020] Preferably, the generation of unified fractal eigenvectors includes:

[0021] The segment feature vectors at each scale are weighted and fused, with the weights dynamically allocated based on the entropy gradient of the behavior within the segment.

[0022] The main coupling patterns are extracted by singular value decomposition based on the inter-scale coupling matrix;

[0023] The weighted segment features are concatenated with the cross-scale coupling vector to generate the final fractal encoding vector.

[0024] Preferably, calculating the regional behavioral entropy and driving adaptive partitioning includes:

[0025] The scenic area is divided into initial semantic units, and the fractal feature vectors of tourists are aggregated in each region;

[0026] Calculate the regional behavioral entropy to characterize the diversity of behaviors within the region;

[0027] An adaptive dynamic partitioning strategy based on behavioral entropy is designed to subdivide high-entropy regions and merge low-entropy regions.

[0028] Preferably, dynamically identified hotspot areas include:

[0029] Based on the dynamically adjusted zoning, and taking into account both the number of tourists and the complexity of their behavior, a hotspot score is generated for each area.

[0030] When the hotspot score exceeds the set threshold, the area is marked as a hotspot area.

[0031] Preferably, constructing a behavior-driven potential field for multi-scale adaptive prediction includes:

[0032] Construct a time-series sequence of tourist traffic for each region;

[0033] By combining hotspot scores and regional aggregation fractal characteristics, the flow potential field between regions is defined;

[0034] By integrating regional flow time series with potential field, an adaptive prediction model is constructed;

[0035] The prediction results are verified and error feedback is provided in real time, and the model parameters are dynamically adjusted.

[0036] Preferably, the adaptive prediction model is implemented using a Long Short-Term Memory (LSTM) network.

[0037] Preferably, generating multi-dimensional candidate strategies and optimizing them through comprehensive scoring and closed-loop methods includes:

[0038] Construct a scheduling objective function to balance tourist density and regional carrying capacity in the scenic area;

[0039] Multiple candidate scheduling strategies are generated, including traffic guidance strategies, regional resource adjustment strategies, and time window peak-shifting strategies;

[0040] Candidate strategies are evaluated in real time, and the optimal strategy is selected using a comprehensive scoring function;

[0041] Collect actual operational data, dynamically update the strategy candidate set and scoring function, and form a closed-loop optimization mechanism.

[0042] The technical solution of this invention: A smart computing center spatiotemporal big data intelligent scheduling and analysis system, which is used to execute the above-mentioned smart computing center spatiotemporal big data intelligent scheduling and analysis method, comprising:

[0043] A multi-source data fractal coding and basic feature construction module is used for data acquisition, cleaning, standardization and fractal feature extraction;

[0044] Entropy-driven dynamic partitioning and hotspot identification module is used to calculate behavioral entropy, dynamic partitioning, and hotspot identification;

[0045] A prediction-driven tourist flow trend extrapolation module is used to construct behavioral potential fields, perform multi-scale predictions, and provide error feedback.

[0046] The intelligent scheduling strategy generation and closed-loop optimization module is used to generate scheduling strategies, evaluate them, and perform closed-loop optimization.

[0047] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects:

[0048] This invention designs a method and system for intelligent scheduling and analysis of spatiotemporal big data in a smart computing center. By fusing multi-source heterogeneous data and semantic trajectory mapping, it effectively improves the accuracy and interpretability of tourist behavior feature extraction. Utilizing entropy-driven adaptive multi-scale feature construction and cross-scale coupling analysis, it comprehensively captures short-term and long-term tourist behavior patterns, significantly enhancing the representation ability of complex spatiotemporal behavior patterns. Through dynamic partitioning of behavioral entropy and hotspot identification mechanisms, it achieves accurate perception and adaptive regional division of scenic area population density and behavioral diversity, improving the sensitivity and reliability of hotspot area identification. Combining behavioral potential fields and multi-scale prediction models improves the accuracy of predicting short- and medium-term tourist flow trends and provides a reliable basis for scheduling decisions. Finally, through a closed-loop optimized intelligent scheduling strategy generation and evaluation mechanism, it achieves efficient allocation of scenic area resources and real-time optimization of tourist distribution, comprehensively improving the scenic area's operational efficiency, resource utilization, and tourist experience. Attached Figure Description

[0049] Figure 1 This is a flowchart of a method for intelligent scheduling and analysis of spatiotemporal big data in a smart computing center, as proposed in this invention.

[0050] Figure 2 This is a system architecture diagram of a spatiotemporal big data intelligent scheduling and analysis system for an intelligent computing center proposed in this invention. Detailed Implementation

[0051] Example 1, as Figure 1 As shown, the present invention proposes a method for intelligent scheduling and analysis of spatiotemporal big data in intelligent computing centers, which includes the following specific implementation steps:

[0052] S1. Through multi-source data normalization and semantic trajectory mapping, short-term and long-term behavioral features are extracted adaptively in multi-scale segments to construct a cross-scale coupling matrix. A unified feature vector is then generated through fractal coding to provide accurate input for the scenic area's time-space temperature control. The specific implementation process is as follows:

[0053] S11. Collect multi-source heterogeneous data within the scenic area, including but not limited to: tourist movement trajectories (GPS location sequence), gate entry and exit records, ticket reservation times, surveillance video counts, crowd heat maps, and weather and event data;

[0054] Because the collected multi-source heterogeneous data is normalized, and each data point is weighted according to the reliability of the data source, signal quality, and consistency of neighborhood information, w p Establish the original trajectory sequence:

[0055]

[0056] Where T represents the set of cleaned and normalized tourist trajectories; p represents the trajectory point number, p = 1, 2, ..., N; N is the total number of trajectory points; t p Indicates the acquisition time of trajectory point p; (x p ,y p ) represents the spatial coordinates of trajectory point p; a p This indicates additional attributes (such as visitor type, ticket category, entrance ID, behavior tags, etc.); w p The confidence weight of the data points is assigned based on the reliability of the data source, signal quality, and neighborhood consistency.

[0057] S12. Map the trajectory to semantic units of the scenic area (such as attractions, passageways, squares, queuing areas). Directly processing coordinate points makes it difficult to understand behavioral patterns. Specifically:

[0058] The scenic area map is divided into M semantic units;

[0059] Calculate the nearest cell ID u for each trajectory point p and projection distance d p :

[0060]

[0061] Among them, u p This represents the semantic unit ID to which trajectory point p belongs, used to map the location to a scenic area's functional zone; Center j d represents the center coordinates of the j-th semantic unit; p This represents the distance from trajectory point p to the center of its semantic unit, used to measure the degree of deviation within the unit;

[0062] Simultaneously, the dwell time and behavior category statistics of each point within each unit are calculated, and a semantic trajectory sequence T is output. s :T s ={(t p ,u p ,a p ,w p ,d p )};

[0063] S13. An entropy-driven adaptive fractal scale selection method dynamically selects the effective scale set for each trajectory, specifically as follows:

[0064] Let the candidate scale set be S0 = {s1, s2, ..., s}. i ,…,s m (e.g., 5, 15, 60, 240 minutes);

[0065] For each scale s, a sliding window segmentation is performed to obtain the segment set D. s ={D s,1 D s,2 ,…};

[0066] Calculate the spatial fluctuation energy and behavioral entropy within the segment:

[0067]

[0068] Among them, E s (T) represents the spatial fluctuation energy at scale s, used to measure the spatial dispersion of the trajectory within a segment; D s Represents the set of trajectory segments at scale s; (x D ,y D Var(x) represents the set of coordinates of all trajectory points within segment D; D ,y D () represents the variance of coordinates within segment D, reflecting the spread or concentration of the trajectory; H represents the average confidence weight within a segment, taking into account the reliability of the data source; s (T) represents the behavioral entropy at scale s, used to quantify behavioral diversity within a segment; p D,k This represents the probability of the k-th type of behavior occurring within segment D;

[0069] Based on the combined energy and entropy scores, an effective scale is selected:

[0070]

[0071] Keep Score s >η (a set threshold, retaining scales with scores higher than η) as S(T);

[0072] Among them, Score s α represents the overall score of scale s, used to determine whether the scale should be retained; E and α H These represent the energy and entropy weighting coefficients, set according to the scenic area management requirements; minE, maxE, minH, and maxH represent the minimum and maximum values ​​of energy and entropy for all candidate scales, used for normalization.

[0073] S14. For each segment D s,i Extract statistical and information-theoretic features, and output segment feature vector G. s :

[0074]

[0075] in, This represents the weighted average speed, reflecting the overall movement speed of tourists within the segment; w p r represents the weight (confidence level) of the trajectory points; stay Indicates the dwell time percentage, measuring the proportion of static behavior within a segment; v p Var represents the instantaneous velocity of trajectory point p; v Represents the weighted velocity variance, used to measure the instability or variability of velocities within segments; 1[v p <v th ] represents the indicator function, when the velocity v of the trajectory point... p Less than threshold v th The value is 1 if the action is active, and 0 otherwise, used to identify dwelling behavior; v th Indicates the speed threshold; This represents the entropy of behavior categories within a segment, used to quantify behavioral diversity. It represents the entropy of the trajectory turning angle distribution within a segment, measuring the complexity of the tourist path;

[0076] S15. Introducing the inter-scale coupling matrix:

[0077] Wherein, SIC(T) represents the inter-scale coupling matrix, which quantifies the correlation of features at different time scales; Representing scales s i s j The next feature summary vector; cov(·,·) represents the covariance function, used to measure the linear correlation of features between scales; r represents the number of effective scales selected;

[0078] S16. The multi-scale segment features and cross-scale coupling features are fused into a unified fractal vector F(t), specifically as follows:

[0079] For each scale s∈S(T), the segment feature vector G s Weighted fusion is performed, with weights dynamically allocated based on the entropy gradient of behaviors within the segment:

[0080] Based on the inter-scale coupling matrix (SIC matrix) SIC(T), the main coupling patterns are extracted through singular value decomposition (SVD) to record the mutual influence between behavioral patterns at different time scales: SIC(T)=UΣV Τ ;

[0081] The weighted segment features are concatenated with the cross-scale coupling vector to generate the final fractal encoding vector:

[0082]

[0083] Among them, w s β represents the weight of scale s, characterizing the importance of that time scale in the overall fractal coding; β represents the entropy sensitivity coefficient, used to adjust the sensitivity of the weight to changes in behavioral entropy. G represents the rate of change of behavioral entropy over time at scale s, i.e., the intensity of short-term behavioral changes at that scale; S(T) represents the set of effective time scales for the current trajectory after entropy-driven selection; G s ves(Σ) represents the feature summary vector at scale s; U and V represent singular vector matrices describing the inter-scale co-structure; Σ represents the singular value vector, where each element represents the strength of a primary coupling mode; k ) indicates that the top k largest singular values ​​are taken to form a cross-scale coupled feature vector, retaining the most significant coupling pattern; F(T) represents the final fractal encoding vector, which uniformly represents the multi-scale behavioral characteristics and cross-scale coupling information of a tourist or region at the current time; This represents a vector concatenation operation, which concatenates the weighted multi-scale feature vector with cross-scale coupled features to form a unified vector.

[0084] S2. By projecting fractal features onto the regional level, calculating regional behavioral entropy and driving adaptive adjustment of partitions, and based on this, combining a comprehensive evaluation of tourist flow and entropy contribution rate, hot spots in scenic areas are dynamically identified, and complex tourist behaviors are flexibly and accurately depicted. The specific implementation process is as follows:

[0085] S21. Based on the scenic area's planning and functional layout, the scenic area is divided into initial semantic units (including but not limited to attractions, passageways, plazas, and rest areas). Within each area, the fractal feature vectors of each tourist from step S1 are aggregated to generate regional-level representation vectors.

[0086] Among them, z l (t) represents the aggregated fractal feature vector of region l at time t, obtained by averaging and aggregating the fractal feature vectors of all tourists in the region. It is used to characterize the overall behavioral characteristics of the region and is the basic representation for behavioral entropy calculation and dynamic partitioning; U l (t) represents the set of tourists belonging to region l at time t; |U l (t)| represents the number of tourists within region l, i.e., set U. l The number of elements in (t); F i (T) represents the fractal feature vector of tourist i;

[0087] S22. Introduce the regional behavioral entropy index to characterize the diversity of behaviors within a region:

[0088]

[0089] Among them, Hl (t) represents the behavioral entropy of region l at time t; p l,k (t) represents the probability of the k-th type of behavior within region l; K represents the total number of behavior categories (including but not limited to staying, sightseeing, passing through, queuing);

[0090] S23. Based on the calculated regional behavior entropy, design an adaptive dynamic partitioning strategy:

[0091]

[0092] Where Δl represents the partitioning adjustment result of region l; θ high Represents the high-entropy threshold; θ low The low-entropy threshold is represented by N(l); N(l) represents the set of adjacent regions of region l; split(l) represents the subdivision operation of the high-entropy region, which is divided according to the spatial and behavioral distribution within the region; merge(l,l') represents merging the low-entropy region with its neighboring regions.

[0093] Accordingly: Through an entropy-driven dynamic partitioning mechanism, the partitioning granularity can adaptively change in real time according to the complexity of tourist behavior; high-density, high-diversity areas are automatically subdivided, enabling the scheduling system to more accurately identify congestion and potential hotspots; low-density, low-complexity areas are merged, reducing the waste of computing resources.

[0094] S24. Based on the dynamically adjusted zoning, and taking into account both visitor numbers and behavioral complexity, generate a hotspot score for each area:

[0095]

[0096] Among them, S l (t) represents the hotspot score of region l at time t; max j |U j (t) represents the maximum number of tourists in all regions at the current moment; max j |H j (t)| represents the maximum behavioral entropy of all regions at the current time; λ1 and λ2 represent the weighting coefficients;

[0097] When S l When (t) exceeds the set threshold τ, the area is marked as a hotspot, triggering subsequent tourist flow prediction and intelligent scheduling measures.

[0098] S3. By fractal features of the region, behavioral entropy, and historical flow sequences, a behavior-driven potential field is constructed. Multi-scale adaptive prediction is then performed by combining the neighborhood potential field with historical sequences. Real-time error feedback is used to dynamically optimize tourist flow trends, thus completing short- and medium-term predictions. The specific implementation process is as follows:

[0099] S31. For each dynamically adjusted region l, construct its tourist flow time series:

[0100] X l (t)=[|U l (tT h )|,|U l (tT h +1)|,...,|U l (t)|];

[0101] Among them, X l (t) represents the historical tourist flow sequence of region l at time t; T h Indicates the length of the history window; |U l (t) represents the number of tourists in region l at time t;

[0102] S32, Combining Hotspot Score l (t) and the fractal characteristics of regional aggregation z l (t), defining the flow potential field between regions:

[0103]

[0104] Where, Φ l→l' (t) represents the potential flow potential from region l to the neighboring region l′; S l' (t) represents the hotspot score of region l′; z l' (t) represents the aggregated fractal eigenvector of region l′; α' and β' represent the weighting coefficients;

[0105] S33, X the regional traffic time sequence l (t) and potential field Φ l→l' (t) Fusion to construct an adaptive prediction model:

[0106]

[0107] in, Represents the predicted number of tourists in region l at future time t+Δt; |U l' (t)| represents the number of tourists in region l′ at time t; f θ (·) indicates that it is based on the historical sequence X l The prediction function of (t) is a Long Short-Term Memory (LSTM) network used in this embodiment; γ represents the weight coefficient.

[0108] S34. Real-time verification and error feedback of prediction results:

[0109]

[0110] The prediction model parameters are dynamically adjusted to make short-term flow forecasts more closely reflect actual tourist behavior.

[0111]

[0112] Where, ε l (t+Δt) represents the prediction error of region l; γ(t+1) represents the weight adjustment value at the next time step; η represents the learning rate, which controls the adjustment magnitude.

[0113] S4. Based on predicted traffic and hotspot behavior characteristics, a scheduling objective function is constructed to generate multi-dimensional candidate strategies. The optimal solution is selected through comprehensive scoring, and combined with real-time feedback closed-loop optimization, the intelligent allocation of scenic area resources and tourist distribution are adaptively adjusted. The specific implementation process is as follows:

[0114] S41. To balance tourist density and regional carrying capacity, define the scheduling optimization objective function:

[0115]

[0116] Wherein, J(t) represents the scheduling objective function value, reflecting the degree of optimization of the overall tourist distribution in the scenic area; C represents the predicted number of tourists in region l at time t; l The maximum carrying capacity of region l is represented by w1 and w2, which represent weighting coefficients.

[0117] S42. Based on the predicted traffic and objective function, generate multiple candidate scheduling strategies, including but not limited to:

[0118] Traffic flow guidance strategy: Adjust scenic area passage signs and attraction guidance signs to guide tourists to move according to the predicted flow distribution;

[0119] Regional resource adjustment strategy: Dynamically allocate resources such as rest areas, restaurants, and restrooms to match the carrying capacity of popular areas with tourist flow;

[0120] Peak-hour strategy: For certain peak areas or attractions, tourists are advised to visit during off-peak hours or make reservations in advance to guide a more balanced distribution of visitors.

[0121] S43. Evaluate candidate strategies in real time using a comprehensive scoring function:

[0122]

[0123] Among them, Q k (t) represents the comprehensive score of candidate strategy k at time t; J k (t) represents the objective function value calculated after strategy k is implemented; Rs k κ1 and κ2 represent the resource utilization efficiency under strategy k; κ1 and κ2 represent the weight coefficients.

[0124] S44. Collect information such as actual tourist flow, congestion in popular areas, and resource utilization rate, calculate the error, and dynamically update the strategy candidate set and scoring function based on the error and real-time feedback to form a continuous optimization closed loop. This ensures that the scheduling strategy adapts to changes in tourist behavior and continuously improves tourist distribution and scenic area operation efficiency.

[0125] Example 2, as Figure 2 As shown, the present invention proposes a smart computing center spatiotemporal big data intelligent scheduling and analysis system, which is used to execute a smart computing center spatiotemporal big data intelligent scheduling and analysis method proposed in Embodiment 1. It includes: a multi-source data fractal encoding and basic feature construction module, an entropy-driven dynamic partitioning and hotspot identification module, a prediction-driven tourist flow trend inference module, and an intelligent scheduling strategy generation and closed-loop optimization module.

[0126] The multi-source data fractal coding and basic feature construction module is responsible for collecting data from various sensors, access control, positioning devices and the environment within the scenic area, cleaning and standardizing the data, and extracting fractal codes and basic features to form spatiotemporal feature vectors that can be used for analysis.

[0127] The entropy-driven dynamic partitioning and hotspot identification module calculates the behavioral entropy of each region, performs dynamic partitioning, identifies hotspot regions and high-risk congested regions, and generates hotspot scores and neighborhood association information to provide input for prediction.

[0128] The prediction-driven tourist flow trend extrapolation module uses zoning features, historical flow sequences, and behavioral potential fields to construct a multi-scale adaptive prediction model, generate short- and medium-term tourist flow trends, and simultaneously provide error feedback to optimize prediction accuracy, providing a basis for scheduling decisions.

[0129] The intelligent scheduling strategy generation and closed-loop optimization module constructs a scheduling objective function based on prediction results and hotspot information, generates multi-dimensional candidate strategies (traffic guidance, resource allocation, and peak-shifting strategies), selects the optimal strategy through strategy evaluation, and performs closed-loop adaptive optimization in combination with real-time traffic and resource usage to form an executable intelligent scheduling scheme.

[0130] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for intelligent scheduling and analysis of space-time big data in a cognitive center, characterized in that, The specific implementation steps include the following: S1, extract the short-term and long-term behavior characteristics of tourists by multi-source data normalization and semantic trajectory mapping, construct a cross-scale coupling matrix and generate a unified fractal feature vector; S2, project the fractal features to the regional level, calculate the regional behavior entropy and drive the adaptive adjustment of the partition, dynamically identify the hot areas by combining the tourist flow and entropy contribution rate, specifically including: According to the scenic spot planning and function layout, the scenic spot is divided into initial semantic units, and in each region, the fractal feature vectors of each tourist are aggregated to generate a region-level representation vector: ; where z l (t) denotes the aggregated fractal feature of region l at time t; U l (t) denotes the set of visitors belonging to region l at time t; |U l (t) denotes the number of visitors in region l, i.e., the cardinality of set U l (t) denotes the number of elements in set U i (T) denotes the fractal feature vector of visitor i; Calculate the regional behavior entropy to describe the diversity of the internal behavior of the region; Design an adaptive dynamic partitioning strategy based on behavior entropy, subdivide high-entropy regions above a set high-entropy threshold, and merge low-entropy regions below a set low-entropy threshold; On the basis of dynamic adjustment of the partition, generate a hot spot score for each region by considering the number of tourists and the complexity of behavior: ； In the formula, S l (t) represents the hotspot score of region i at time t; max j |U j (t) represents the maximum number of tourists of all regions at the current time; max j |H j (t) represents the maximum behavior entropy of all regions at the current time; λ1 and λ2 represent weight coefficients; H l (t) represents the behavior entropy of region i at time t; When the hot spot score exceeds a set threshold, mark the region as a hot spot region; S3, fuse the regional fractal features, behavior entropy and historical flow sequence, construct a behavior-driven potential field, and perform multi-scale adaptive prediction and error feedback optimization, including: Construct a time series sequence of tourist flow in each region; Combine the hot spot score and the regional aggregated fractal features to define the flow potential field between regions; Fuse the regional flow time series and the potential field to construct an adaptive prediction model; Real-time check and error feedback on the prediction results to dynamically adjust the model parameters; S4, based on the prediction results, construct a scheduling objective function, generate a multi-dimensional candidate strategy, and through comprehensive scoring and closed-loop optimization, intelligently schedule the scenic resources and optimize the distribution of tourists. 2.The method of claim 1, wherein, Multi-source data normalization and semantic trajectory mapping includes: Collecting multi-source heterogeneous data in the scenic area, including tourist movement trajectories, gate entry and exit records, ticket reservation times, monitoring video counts, crowd heat maps, weather and event data; Clean and normalize the data, and assign a confidence weight to each data point; Map the trajectory points to semantic units, calculate the dwell time and behavior category statistics of each unit, and output the semantic trajectory sequence. 3.The method of claim 2, wherein, Constructing a cross-scale coupling matrix and generating a unified fractal feature vector includes: An entropy-driven adaptive fractal scale selection method is used to dynamically select the effective scale set for each trajectory; Extract statistical features and information theory features for each scale of the trajectory segment, and output the segment feature vector; Construct a scale coupling matrix to quantify the correlation of features at different time scales; Fuse the multi-scale segment features and cross-scale coupling features into a unified fractal feature vector. 4.The method of claim 3, wherein, The generation of the unified fractal feature vector includes: Weighted fusion of segment feature vectors at each scale, with weights dynamically allocated based on the behavior entropy gradient within the segment; Extract the main coupling mode based on the inter-scale coupling matrix through singular value decomposition; Concatenate the weighted segment features and the cross-scale coupling vector to generate the final fractal encoding vector. 5.The method of claim 4, wherein, The adaptive prediction model is implemented using a long short-term memory network (LSTM). 6.The method of claim 5, wherein, Generating multi-dimensional candidate strategies and through comprehensive scoring and closed-loop optimization includes: Construct a scheduling objective function to balance the scenic tourist density and regional carrying capacity; Generate multiple candidate scheduling strategies, including flow guidance strategies, regional resource adjustment strategies and time window peak shifting strategies; Real-time evaluation of candidate strategies, using a comprehensive scoring function to select the optimal strategy; Collecting actual operation data, dynamically updating the strategy candidate set and scoring function, forming a closed-loop optimization mechanism.

7. The intelligent center space-time big data intelligent scheduling analysis system for performing the intelligent center space-time big data intelligent scheduling analysis method of any one of claims 1-6. Comprise: Multi-source data fractal encoding and basic feature construction module for data collection, cleaning, standardization and fractal feature extraction; Entropy-driven dynamic partitioning and hotspot identification module for calculating behavior entropy, dynamic partitioning and hotspot identification; Prediction-driven tourist flow trend deduction module for building behavior potential field, multi-scale prediction and error feedback; Intelligent scheduling strategy generation and closed-loop optimization module for generating scheduling strategies, evaluation and closed-loop optimization.

Images