A cloud computing and big data based smart park collaborative service system

Abstract

The application discloses a kind of wisdom park collaborative service systems based on cloud computing and big data, it is related to wisdom park technical field. Including data security and privacy protection module, it is characterized by: data acquisition module: for real-time acquisition wisdom park multi-source data;Data processing module: for building data fusion and collaborative model, the feature extraction and fusion of wisdom park multi-source data are carried out, then based on big data, collaborative analysis is carried out, and collaborative decision suggestion is generated, and data intercommunication sharing is realized;Wisdom service and optimization module: for building wisdom service model, based on service request, automatically generate service, and real-time track and feedback service progress, optimize and update model, improve service response efficiency.The application builds wisdom service model, real-time analysis multi-source service request, automatically generate service, and optimize and update model, continuously improve service response speed and user service experience, and reduce user loss risk.

Description

Technical Field

[0001] This invention relates to the field of smart park technology, specifically to a smart park collaborative service system based on cloud computing and big data. Background Technology

[0002] The smart park service system is a key vehicle for the transformation and upgrading of traditional parks in the digital economy era. It originates from the need to solve practical pain points such as park management efficiency, service experience, cost control, and security. Relying on information technologies such as the Internet of Things, cloud computing, big data, and artificial intelligence, it aims to create a modern park ecosystem that is highly efficient in management, convenient in services, pleasant in environment, active in industry, and green and sustainable.

[0003] For example, patent publication number CN115423664A, entitled "A Smart Park IoT Integrated Service System," includes: an energy management system, an access control system, a video surveillance system, a fire alarm system, an image processing system, an intelligent cleaning system, a processor, and a display system; all systems are uniformly connected by the processor. This invention enables a high degree of intelligence in the park and facilitates centralized management.

[0004] However, the above-mentioned and similar technical solutions lack in-depth data analysis, which may lead to incomplete decision-making recommendations, delayed resource allocation, and slow response speed to user service requests, resulting in poor user experience, reduced user satisfaction, and increased risk of user churn. Summary of the Invention

[0005] The purpose of this invention is to provide a smart park collaborative service system based on cloud computing and big data to solve the problems mentioned in the background.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a smart park collaborative service system based on cloud computing and big data, including a data security and privacy protection module, characterized in that:

[0007] Data acquisition module: used for real-time acquisition of multi-source data in the smart park;

[0008] Data processing module: used to build data fusion and collaboration models, extract and fuse features from multi-source data in smart parks, and then perform collaborative analysis based on big data to generate collaborative decision-making suggestions and achieve data interoperability and sharing;

[0009] Intelligent Service and Optimization Module: Used to build intelligent service models, automatically generate services based on service requests, track and provide feedback on service progress in real time, optimize and update models, and improve service response efficiency.

[0010] Furthermore, the method for creating the data fusion and collaboration model includes:

[0011] Data processing: The collected multi-source data of the smart park is standardized and cleaned, and coordinate transformation and timestamp synchronization are performed according to a unified spatiotemporal benchmark to obtain the processed multi-source data of the smart park;

[0012] Feature fusion: Based on the spatiotemporal-semantic association knowledge network, cross-modal feature association is performed on the processed multi-source data of the smart park to establish a fusion database;

[0013] Collaborative Decision Making: Based on big data, analyze and integrate data from the database, explore the relationships between data, and generate collaborative decision-making suggestions.

[0014] Furthermore, the method for constructing the spatiotemporal-semantic association knowledge network includes:

[0015] Data labeling: Assigning preliminary semantic labels to the processed multi-source data of the smart park;

[0016] Geospatial Relationship Construction: Based on the processed multi-source data of the smart park, a spatial geographic model of the smart park is constructed; spatial relationships between different spatial elements are calculated, including: distance, adjacency, and containment;

[0017] Time element extraction: After analyzing and processing the multi-source data of the smart park, extract time patterns, mine time patterns, and obtain spatiotemporal elements;

[0018] Semantic knowledge extraction and fusion: Extract semantic knowledge of smart parks from multi-source data and relevant literature on smart parks after processing, and construct a semantic knowledge base; based on the semantic knowledge base, link the multi-source data and semantic knowledge of smart parks after processing.

[0019] Network Construction: Using processed multi-source data of the smart park, spatiotemporal elements, semantic knowledge, and spatial geographic models and their spatial relationships as nodes, and the associations between nodes and the weights of spatial relationships as edges, a spatiotemporal-semantic association knowledge network is constructed.

[0020] Furthermore, the step of generating collaborative decision recommendations includes:

[0021] Data analysis: Perform multi-dimensional analysis on the data in the integrated database, including time dimension, spatial dimension, and modal dimension, to obtain data analysis results;

[0022] Association rule mining: Based on the data analysis results, the relationships between data in the integrated database are mined; and based on the relationships, future data trends and events are predicted to obtain prediction results;

[0023] Generate decision recommendations: Construct decision scenarios, determine decision indicators and influencing factors; and generate collaborative decision recommendations for the corresponding scenarios based on correlations and prediction results.

[0024] Furthermore, the feature fusion method includes:

[0025] Cross-modal feature association: The processed smart park data is spliced ​​to generate a basic feature vector; multimodal features are weighted and aggregated to obtain a multimodal feature vector;

[0026] Dimensionality reduction and optimization: The spatiotemporal-semantic association knowledge network and multimodal feature vectors are dimensionality reduced to obtain dimensionality-reduced feature vectors, and the feature weights are adjusted to establish a fusion database.

[0027] Furthermore, the method for cross-modal feature association includes:

[0028] Concatenation: Connect the feature vectors of different modalities aligned within the same spatiotemporal unit to obtain the basic feature vector;

[0029] Calculate feature weights: Automatically learn the specific content of the basic feature vectors, obtain the dynamic weight of each feature dimension, and highlight key information;

[0030] Weighted aggregation: The calculated feature weights are summed with the basic feature vectors to obtain the multimodal feature vector.

[0031] Furthermore, the method for adjusting feature weights includes:

[0032] Normalized dimensionality reduction features: The dimensionality-reduced feature vectors are standardized, and the initial weights are redistributed based on the contribution of feature variance.

[0033] Calculate dynamic weights: Analyze the semantic relationships between multimodal features and calculate dynamic weight coefficients;

[0034] Similarity-driven adjustment: Calculate the similarity matrix between feature vectors and adjust the feature weights according to the strength of the correlation.

[0035] Furthermore, the method for creating the intelligent service model includes:

[0036] Service Function Integration: Integrate various service function modules;

[0037] Process definition: Define the processing flow for various services, including task allocation, processing time limits, and workflow paths;

[0038] Information input: The user inputs a service request;

[0039] Service execution: Based on the processing flow, the service type is determined according to the service request, the service is automatically generated, and the service is transferred to the relevant service.

[0040] Service feedback and optimization: Track the progress of service request processing in real time and provide feedback to users; after the service is completed, push a service evaluation survey to users, analyze the collected service evaluation data, and optimize the processing flow based on the analysis results.

[0041] Furthermore, the service execution steps include:

[0042] Determine service type: Parse the user's service request, identify the service type, and match it with a predefined processing flow;

[0043] Service generation: According to the processing flow, service tasks are assigned to the corresponding processing nodes, and processing progress data is synchronized to the fusion database in real time.

[0044] Furthermore, the service feedback and optimization includes handling service timeouts and rule conflicts. The handling methods include: automatically triggering an early warning mechanism and transferring the call to a manual review node, notifying the user of the reasons for the service timeout and the solution, and updating the fusion database.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] A smart park collaborative service system based on cloud computing and big data integrates multi-source data by constructing a data fusion and collaboration model. Based on big data, it performs in-depth analysis and prediction of the data, grasps the real-time status and usage patterns of various resources in the park, and enables the rational allocation of park resources. This achieves a leap from data interconnection to intelligent decision-making and improves the accuracy of park management.

[0047] Meanwhile, by building a smart service model, we can analyze multi-source service requests in real time, automatically generate services, and optimize and update the model to continuously improve service response speed and user service experience, as well as reduce the risk of user churn. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the system modules of the present invention;

[0049] Figure 2 This is a schematic diagram of the data fusion and collaboration model of the present invention;

[0050] Figure 3 This is a schematic diagram of the spatiotemporal-semantic association knowledge network construction method of the present invention;

[0051] Figure 4 This is the intelligent service model of the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, the present invention provides a technical solution: a smart park collaborative service system based on cloud computing and big data, comprising:

[0054] Data acquisition module: used for real-time acquisition of multi-source data in the smart park;

[0055] It is important to note that the multi-source data collected for the smart park includes: basic user data, facility and equipment data, service and feedback data, vehicle data, and enterprise data, which are transmitted to the cloud data center in real time. Basic user data is obtained through facial / fingerprint recognition machines, mobile apps / mini-programs, and WiFi probes; facility and equipment data is obtained through IoT sensors, PLC systems, equipment manufacturer APIs, and smart meters / water meters; service and feedback data is obtained through mobile devices, smart terminals, and API gateway logs; vehicle data is obtained through entrance and exit cameras, geomagnetic sensors, GPS positioning, and ETC; and enterprise data is obtained through API integration with enterprise OA systems.

[0056] Data processing module: used to build data fusion and collaboration models, extract and fuse features from multi-source data in smart parks, and then perform collaborative analysis based on big data to generate collaborative decision-making suggestions and achieve data interoperability and sharing;

[0057] like Figure 2 As shown, this invention provides a data fusion and collaboration model;

[0058] Specifically:

[0059] H1: Data Processing: Standardize and clean the collected multi-source data of the smart park, and perform coordinate transformation and timestamp synchronization according to a unified spatiotemporal benchmark to obtain the processed multi-source data of the smart park.

[0060] H2: Feature Fusion: Based on the spatiotemporal-semantic association knowledge network, the processed multi-source data of the smart park are cross-modal feature associations to establish a fusion database;

[0061] H3: Collaborative Decision Making: Based on big data, analyze and integrate data from the database, mine the relationships between data, and generate collaborative decision-making suggestions;

[0062] The feature fusion method includes: cross-modal feature association: concatenating the processed smart park data to generate a basic feature vector; weighted aggregation of multimodal features to obtain a multimodal feature vector; dimensionality reduction and optimization: dimensionality reduction processing of the spatiotemporal-semantic association knowledge network and the multimodal feature vector to obtain a dimensionality-reduced feature vector, adjusting the feature weights, and establishing a fusion database.

[0063] The specific method for cross-modal feature association includes: concatenation, where feature vectors of different modalities aligned within the same spatiotemporal unit are concatenated end-to-end to obtain a basic feature vector; calculation of feature weights, where the specific content of the basic feature vector is automatically learned to obtain the dynamic weight of each feature dimension, highlighting key information; and weighted aggregation, where the calculated feature weights are weighted and summed with the basic feature vector to obtain a multimodal feature vector.

[0064] The method for adjusting feature weights includes: normalizing and reducing dimensionality features by standardizing the reduced feature vectors and redistributing initial weights based on the contribution of feature variance; calculating dynamic weights by analyzing the semantic correlation between multimodal features and calculating dynamic weight coefficients; and similarity-driven adjustment by calculating the similarity matrix between feature vectors and adjusting feature weights based on the strength of the correlation.

[0065] The steps for generating collaborative decision-making suggestions include: data analysis: performing multi-dimensional analysis on the data in the fusion database, including time, space, and modality dimensions, to obtain data analysis results; association rule mining: mining the association relationships between data in the fusion database based on the data analysis results; and predicting future data trends and events based on the association relationships to obtain prediction results; and generating decision-making suggestions: constructing decision-making scenarios, determining decision indicators and influencing factors; and generating collaborative decision-making suggestions for the corresponding scenarios based on the association relationships and prediction results.

[0066] It is important to note the following data processing steps: ETL tools are used to standardize and clean the collected multi-source data from the smart park; outliers are automatically corrected based on thresholds set according to park business rules; missing values ​​are imputed using the KNN algorithm; and a unified field definition is established for the master data management module, such as standardizing customer names to Unicode encoding. The local coordinate system is converted to the WGS84 standard using Gauss-Kruger projection, with the error controlled within ±0.5 meters; finally, an NTP server cluster is deployed to ensure timestamp deviation is less than 50ms.

[0067] Feature fusion: First, cross-modal feature association is performed: feature concatenation, which joins different modal feature vectors aligned within the same spatiotemporal unit, such as CNN features extracted from video and sensor time-series data, to form a basic feature vector; the `torch.cat` function in PyTorch is used to merge the vectors. Note that before concatenation, it is crucial to ensure that the feature vectors from different modalities are spatiotemporally aligned and have the same dimensionality (all feature vectors are standardized to the same length). Next, feature weights are calculated using a cross-modal attention mechanism, automatically assigning weights through a learnable parameter matrix, such as using a multi-head self-attention layer to calculate the dynamic weight coefficients for each feature dimension, highlighting key information, such as features related to energy consumption peaks. Finally, weighted aggregation is performed using PyTorch's broadcast mechanism to sum the dynamic weights with the basic feature vector, generating a multimodal feature vector. Note that the weights must be normalized to avoid bias caused by differences in feature scale.

[0068] Next, dimensionality reduction and optimization are performed: Dimensionality reduction is achieved by decomposing eigenvalues ​​through principal component analysis (PCA) of the spatiotemporal semantic association knowledge network and multimodal feature vectors stored in the Neo4j graph database, retaining principal components with 95% variance contribution. Note that the feature dimension after dimensionality reduction needs to be less than 30% of the original data to ensure an information loss rate ≤5%. Dimensionality-reduced features are normalized using Z-score standardization, and initial weights are redistributed based on feature variance contribution; higher variance results in higher feature weights. Dynamic weights are calculated by analyzing the semantic associations between multimodal feature vectors, such as device-energy consumption relationships, using GAT, and outputting dynamic weight coefficients; then, node correlations are calculated using GAT. Similarity-driven adjustment is performed by calculating the cosine similarity matrix between feature vectors and adjusting weights based on association strength. Weight adjustment methods include: when similarity is greater than 0.7, feature weights are superimposed; when similarity is between 0.3 and 0.7, conditional weighting is used; when similarity is less than 0.3, weighted filtering is applied.

[0069] Finally, the fusion database is established: It is a heterogeneous database based on feature-layer fusion, supporting hierarchical storage: hot data is stored in Redis; warm data is stored in the InfluxDB time-series database; and cold data is archived to AWS S3. The hot data consists of real-time features, the warm data consists of dimensionality-reduced features, and the cold data consists of original multi-source data from the smart park. The data structure uses a graph structure to store spatiotemporal-semantic relationships, such as the entity relationship "camera-person density-BIM area," and feature vectors are stored using Float32, supporting efficient similarity queries. Functionality includes a unified RESTful API interface supporting real-time access from multiple business systems; it also features a dynamic weighted indexing mechanism to accelerate key feature retrieval. This fusion database, through cross-modal association and optimized dimensionality reduction, solves the problem of isolated multi-source data in smart parks, providing high-density feature support for decision-making collaboration.

[0070] Decision Collaboration: Data analysis, including: Time dimension analysis, using STL decomposition and the Prophe model to identify the periodic patterns and trends in the data, and using an LSTM network to predict the data change trend for the next 72 hours, with time alignment accuracy requiring millisecond-level precision; outliers are repaired using cubic spline interpolation with a sliding window size of 1 hour; Spatial dimension analysis, constructing a spatial index based on Geohash encoding, combined with the DBSCAN algorithm, to identify areas with dense equipment and hotspots of personnel activity; GIS data projection uniformly adopts the CGCS2000 coordinate system; Modal dimension analysis, using a multimodal Transformer model for feature alignment, calculating feature similarity through a cross-attention mechanism, with a threshold ≥0.7. Association rule mining, using the FP-Growth algorithm to mine frequent itemsets, such as "peak hours → elevator congestion + energy consumption surge". The PC algorithm distinguishes between correlation and causality, eliminating spurious associations; then, the ST-GNN model fuses graph convolution and LSTM to predict the probability of event occurrence; the prediction results need to be validated through Bootstrap sampling, with a confidence level ≥90%; association rules are generated based on the Apriori algorithm and updated in real time to the Neo4j graph database. The method generates decision recommendations by loading BIM and IoT data using a digital twin engine; determining indicator weights through the Analytic Hierarchy Process (AHP); simulating the effects of different strategies based on the PPO algorithm; and outputting the Pareto optimal solution. Emergency response times are less than 3 seconds, and resource allocation decisions must include three alternative solutions. This method achieves intelligent transformation from data to action through a closed-loop process of multi-dimensional analysis, correlation mining, and dynamic decision-making.

[0071] like Figure 3 As shown, this invention provides a method for constructing a spatiotemporal-semantic related knowledge network;

[0072] Specifically:

[0073] Data labeling: Assigning preliminary semantic labels to the processed multi-source data of the smart park;

[0074] Geospatial Relationship Construction: Based on the processed multi-source data of the smart park, a spatial geographic model of the smart park is constructed; spatial relationships between different spatial elements are calculated, including: distance, adjacency, and containment;

[0075] Spatiotemporal element extraction: After analyzing and processing multi-source data of the smart park, extract time patterns, mine time patterns, and obtain spatiotemporal elements;

[0076] Semantic knowledge extraction and fusion: Extract semantic knowledge of smart parks from multi-source data and relevant literature on smart parks after processing, and construct a semantic knowledge base; based on the semantic knowledge base, link the multi-source data and semantic knowledge of smart parks after processing.

[0077] Network Construction: Using the processed multi-source data of the smart park, spatiotemporal elements, semantic knowledge, and spatial geographic models and their spatial relationships as nodes, and the association relationships between nodes and the weights of spatial relationships as edges, a spatiotemporal-semantic association knowledge network is constructed.

[0078] It is important to note that for text-based data, such as park announcements and equipment maintenance records, word segmentation tools (such as Jieba) are used to break down sentences, and part-of-speech tagging (such as NLTK's POS Tagger) is used to identify key entities, such as "elevator in Building 3" and "fire hydrant A." Semantic understanding is based on the BERT model to label the data category, such as "equipment malfunction" and "personnel access." For video / image data, the YOLO object detection algorithm is used to identify entities, such as pedestrians, vehicles, and fire lanes, and their attributes are labeled, such as "pedestrian - employee" and "vehicle - external vehicle." Based on preset business rules, such as park management regulations, structured data is automatically labeled semantically, such as "card swipe time + personnel ID" in access control records, automatically labeled as "employee Zhang San + 8:40 entered the park." Labeling rules follow an entity-attribute-value model, such as "equipment - status - normal" and "area - function - office area." A unified semantic tag dictionary ensures consistency in labels for the same type of entity / relationship, such as not mixing adjacent areas with neighboring areas.

[0079] Geospatial association construction involves building a 3D spatial model of the smart park based on processed multi-source data. ArcGIS tools are used to digitally model spatial elements such as building coordinates, road directions, and equipment installation locations. The straight-line / actual path distance between two spatial elements is calculated using the Euclidean distance formula. Topological analysis algorithms determine whether elements share boundaries, such as the adjacency of Building 2 and Building 3, or the adjacency of a fire lane and Staircase 2. Spatial inclusion algorithms determine element hierarchies, such as meeting room A being contained within Building 3, or a parking lot containing 120 parking spaces. Spatial relationship rules follow topological theory, namely the topological rules for points, lines, and surfaces, such as surface-to-surface elements not overlapping (except in special areas), and passageways must be adjacent to at least two areas. These rules are combined with the actual layout rules of the park, such as equipment needing to be located within its designated area.

[0080] Spatiotemporal element extraction involves extracting temporal features, such as time points, time periods, and periodic patterns, from the processed multi-source data of the smart park, and associating them with spatial entities to form spatiotemporal elements. Specifically, this includes: using a sliding window algorithm to extract time points and time periods from time-series data such as sensor and access control data; mining patterns using an LSTM time series model, such as the requirement for equipment A to be maintained every 30 days; and using the spaCy time parser to extract timestamps from unstructured data, such as equipment maintenance record text, and associating them with corresponding spatial entities. Time rules: Time granularity is standardized, using a unified "year-month-day-hour-minute-second" format to distinguish between absolute and relative time; and categorized by periodicity into "one-off events," "periodic events," and "trend events." Absolute time refers to a fixed time coordinate point based on the Gregorian calendar, representing the actual physical time of the event's occurrence; relative time refers to a time offset based on a reference event, representing the temporal relationship between events rather than specific moments. By strictly distinguishing between absolute and relative time, the system can simultaneously support precise time point location (the instant of equipment failure) and complex time sequence reasoning (such as "the backup generator will automatically start 30 minutes after a power outage").

[0081] Semantic knowledge extraction and fusion involves extracting domain knowledge, such as equipment functions, management rules, and causal relationships of events, from processed multi-source data and relevant literature of the smart park, and associating them with spatiotemporal elements to form a structured semantic knowledge base. Specifically, this includes: extracting domain entities from relevant literature based on the NER model, such as smart meters, security systems, and emergency plans; extracting semantic relationships between entities using remote monitoring algorithms, such as "smart meter - monitoring - electricity consumption" and "security system - association - camera"; then extracting events to identify key events and elements, such as "equipment failure event, subject: elevator A, time: 2025-07-24-15-30-20, cause: motor damage"; resolving entity ambiguity based on cosine similarity calculation using word vector similarity, such as "Building 3" and "Three Buildings" being the same entity; and fusing multi-source semantic knowledge into a unified knowledge base using Neo4j to avoid duplication or conflict. Semantic association rules: Based on the domain ontology, define semantic relationships, such as "device-belongs to-region" and "event-trigger-response measures"; follow logical reasoning rules, such as "if device A belongs to region B, and region B is contained in park C, then device A belongs to park C".

[0082] The network construction uses processed multi-source data of the smart park, spatiotemporal elements, semantic knowledge, and spatial geographic models and their spatial relationships as nodes, and the associations between nodes and the weights of spatial relationships as edges, to construct a spatiotemporal-semantic association knowledge network. Specifically, this includes: storing nodes and edges as a graph structure using the Neo4j graph database; assigning weights to edges to quantify the strength of associations, such as giving a higher weight to "device A and device B belong to the same maintenance group" than "device A and device C are far apart". Network construction rules include: node uniqueness, with only one node retained for each entity / element in the network, ensured through ID mapping; and standardized relationship types for ease of subsequent querying and reasoning, such as "space-distance", "time-sequence", and "semantic-causal". The weighting of edges uses Euclidean distance to measure the similarity of node attributes, which is then directly used as the edge weight value. The constructed spatiotemporal-semantic association knowledge network is stored in a graph database, such as Neo4j.

[0083] Intelligent Service and Optimization Module: Used to build intelligent service models, automatically generate services based on service requests, track and provide feedback on service progress in real time, optimize and update models, and improve service response efficiency.

[0084] like Figure 4 As shown, this invention provides a smart service model;

[0085] Specifically:

[0086] S1: Service Function Integration: Integrating various service function modules;

[0087] S2: Process Definition: Defines the processing flow for various services, including task allocation, processing time limits, and workflow paths;

[0088] S3: Information Input: The user inputs a service request;

[0089] S4: Service Execution: Based on the processing flow, determine the service type according to the service request, automatically generate a service, and transfer it to the relevant service;

[0090] S5: Service Feedback and Optimization: Track the progress of service request processing in real time and provide feedback to users; after the service is completed, push a service evaluation survey to users, analyze the collected service evaluation data, and optimize the processing flow based on the analysis results.

[0091] The service execution steps include: determining the service type: parsing the service request input by the user, identifying the service type, and matching it with a predefined processing flow; service generation: according to the processing flow, assigning the service task to the corresponding processing node, and synchronizing the processing progress data to the fusion database in real time.

[0092] The service feedback and optimization includes handling service timeouts and rule conflicts. The handling methods include: automatically triggering an early warning mechanism and transferring the call to a manual review node, notifying the user of the reasons for the service timeout and the solution, and updating the fusion database.

[0093] It's important to note the service function integration: This integrates the system with apps / mini-programs / smart terminals, using micro-frontend technology to achieve modular function loading. Users can access all services through single sign-on. Service requests are received and categorized. Requests can enter the smart service system through various channels, such as apps, mini-programs, and self-service terminals. The system automatically records basic information about service requests, including request time, request source, requester identity information, and contact information. Based on the content and nature of the service requests, NLP technology is used to automatically categorize them. For example, security services include requests to view surveillance videos, report access control malfunctions, and security time reports; energy services include reports of power outages, water usage anomalies, and energy usage inquiries; property services include requests for building repairs, complaints about public area cleaning, and suggestions for green space maintenance; and enterprise services include inquiries about enterprise onboarding procedures, service request change requests, and assistance with enterprise event organization. Service requests that cannot be accurately categorized automatically are marked as "awaiting manual categorization" and assigned to customer service personnel for manual judgment and categorization.

[0094] Process Definition and Task Allocation Rules: Allocation based on service type: Security services are assigned to security management departments or relevant technical personnel. For example, requests to view surveillance videos are assigned to monitoring center staff, and access control system malfunction reports are assigned to access control system maintenance personnel. Energy services: Power outage reports are assigned to the power maintenance team, water usage anomaly reports are assigned to the water management department, and energy usage consultations are assigned to energy service specialists. Property services: Building repair requests are assigned to maintenance teams, public area cleaning complaints are assigned to cleaning team leaders, and green space maintenance suggestions are assigned to the green space maintenance department. Enterprise services: Consultations on enterprise onboarding procedures are assigned to enterprise service specialists, service request change requests are assigned to the corresponding service managers, and enterprise event organization assistance is assigned to the event planning team. Allocation based on geographic location: For service requests involving specific areas within the park, such as air conditioning malfunction reports in an office building, the task is assigned to the relevant service team and staff responsible for that area based on the location information provided in the service request. The system can automatically determine the task allocation recipients by combining the park's map information and the jurisdiction of each service team. Based on workload allocation: The system monitors the task load of each service team and staff in real time, including the number of currently unfinished tasks, the urgency level of the tasks, and the estimated completion time. For new service requests, under the premise of meeting the above requirements based on service type and geographical location, the system prioritizes the allocation to service teams or staff with relatively light workloads, so as to achieve balanced task distribution and improve overall service efficiency.

[0095] Special rule allocation: For high-priority service requests, such as those involving security risks, major facility failures, or service needs of important customers, the system adopts priority allocation rules to immediately assign the task to senior technical personnel or dedicated emergency response teams in the relevant service teams and mark it as an urgent task, requiring priority handling; for users with specific service requirements or historical service records, such as long-term cooperative corporate customers or park users with special needs, the system will assign the task to dedicated staff familiar with the customer based on preset customer preferences and service agreements, in order to improve personalized and high-quality service.

[0096] Processing Time Limits: Basic processing time limits are set according to different service types. For example, for security services, requests to view surveillance videos should be responded to within 15 minutes, and video viewing services should be provided; access control malfunction reports should be addressed on-site within 30 minutes; and security incident reports should initiate emergency response procedures within 10 minutes. For energy services, power outage reports should be restored within 3 hours, with specific time limits set based on the severity of the fault and the difficulty of repair; abnormal water usage reports should have the cause identified and preliminary measures taken within 2 hours; and energy usage inquiries should be answered within 20 minutes. For property management services, different processing time limits are set for building repair requests based on the urgency and complexity of the repair. For example, leaky roof repairs should be addressed on-site within 1 hour, wall peeling repairs should be completed within 3 working days, and public area cleaning complaints should be arranged for cleaning within 1 hour. For enterprise services, consultations regarding enterprise onboarding procedures should be provided with detailed procedures and answers within 1 business day; service request change requests should be communicated to the enterprise and feedback should be provided within 3 business days; and assistance with enterprise event organization should begin 5 business days before the event to ensure the smooth running of the event.

[0097] Flow Path Rules: Initial Processing Path: Service requests are assigned to the corresponding service teams and staff according to the above task allocation rules, entering the initial processing stage. Staff need to respond to the service request within the processing time limit, confirm receipt of the task, and begin processing. Staff need to take appropriate measures based on the specific content of the service request, such as conducting on-site inspections, troubleshooting, information retrieval, and communication coordination, and record the processing process and progress in the system. Internal Flow Path: During the processing, if it is found that cooperation from other departments is needed, staff can initiate an internal flow request in the system to transfer the task to the relevant support department or team. For example, if it is found that the equipment to be replaced needs to be modified for energy system design during equipment maintenance, the maintenance personnel will initiate an internal flow request in the system to transfer the task to the energy management department for collaborative processing. After receiving the flow task, the support department or team needs to continue processing according to the prescribed processing time limit and provide feedback on the processing results to the original staff. The original staff will continue to advance the task based on the feedback results to ensure that the service request is resolved as a whole. For external collaboration requests requiring support from companies outside the park, such as the procurement of special materials, the system supports task routing to external collaborators. Staff send detailed collaboration requirements to these external collaborators through the system and track their progress. After completing the task, the external collaborator returns the results to the system, which staff then review and confirm. If the results meet the requirements, the task is marked as completed; otherwise, staff communicate and negotiate with the external collaborator to determine further solutions and continue tracking progress until the task is completed.

[0098] Service execution: When a user inputs and submits a service request through a given channel, the system uses NLP technology to parse the user's input service request, identify the service type, match it with a predefined processing flow, assign the service task to the corresponding processing node, and synchronize the processing progress data to the fusion database in real time.

[0099] Service Feedback and Optimization: The system uses Apache Kafka and Apache Real-time to collect task allocation status, processing progress, and time limits. It then uses an LSTM model to predict process deviations and provides timely feedback to users and administrators. After service completion, staff must fill in detailed service request resolution information, measures taken, and processing time points in the system and submit it to the user for feedback. Simultaneously, the system triggers an email / SMS API via a RabbitMQ message queue, sending users a service evaluation questionnaire to express their satisfaction and opinions. The system analyzes user feedback text using a BERT model, identifying negative emotions such as "waited too long." Dynamic decision-making is based on the DQN algorithm, automatically returning simple issues to the original staff member and escalating complex issues to supervisors until the user is satisfied, achieving closed-loop management of the service process. The system performs monthly statistical analysis of various data during the service request processing process, such as trends in the number of different types of service requests, average processing time, efficiency of each stage, and distribution of user satisfaction ratings. Monthly analysis reports are automatically generated using Python and Matplotlib, and optimization suggestions are generated based on NLP technology to specifically optimize and adjust the service request processing flow. For example, optimize task allocation rules to improve the accuracy and rationality of task allocation; adjust processing time limits to better meet user needs and actual work conditions; and simplify workflow processes to reduce unnecessary time costs. Regularly evaluate the effectiveness of process optimization measures, such as comparing key performance indicators (KPIs) before and after optimization, including processing time limits and customer satisfaction, to verify the effectiveness of the optimization measures. If the expected optimization results are not achieved, analyze the reasons and further adjust the optimization strategy to continuously improve the service request processing flow and enhance the overall performance and user experience of the intelligent service model.

[0100] The service feedback and optimization also includes handling service timeouts and rule conflicts. When a service timeout occurs, an early warning mechanism is automatically triggered, and the task is escalated to the superior management department or complaint handling department. For timeouts caused by special reasons, staff need to submit an application including the reason for the delay and the estimated completion time; after approval, the processing time limit will be adjusted. Simultaneously, the system will push timeout explanations and solutions to users in real time, and all processing status changes will be promptly synchronized to the integrated database. In the case of rule conflicts, the system automatically freezes the current process and transfers it to a manual review node, ensuring that the entire process of anomaly handling is traceable.

[0101] Data security and privacy protection module: used for data encryption and desensitization, access control, and real-time monitoring and recording of data operation behavior to protect data security and user privacy;

[0102] It is important to note that sensitive data stored in the cloud data center is encrypted using the AES encryption algorithm. During data transmission, an encrypted channel is established using the SSL / TLS protocol. An RBAC model is employed, assigning different roles based on the responsibilities and permissions of users within the campus, with each role having corresponding data access permissions. Multiple authentication technologies, such as dynamic passwords and biometrics, are applied to ensure the authenticity of user identities. When analyzing service requests, the data fields requiring anonymization and the degree of anonymization are determined; for example, partial masking is used for name fields, and replacement is used for ID number fields. Anonymized data is applied in data mining and analysis, and data sharing to ensure the privacy and security of user data. A data leakage prevention system is deployed to monitor data access, modification, and transmission in real time; security policies and alarm thresholds are set, and when abnormal data access behavior is detected, timely warnings are issued and corresponding measures are taken to intercept and handle the situation. A data audit system is established to record detailed information on all data operations, including the user, time, type, and target of the operation. Audit logs are regularly analyzed and reviewed to promptly identify potential security issues and risks.

[0103] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. A smart park collaborative service system based on cloud computing and big data, including a data security and privacy protection module, characterized in that: Data acquisition module: used for real-time acquisition of multi-source data in the smart park; Data processing module: used to build data fusion and collaboration models, extract and fuse features from multi-source data in smart parks, and then perform collaborative analysis based on big data to generate collaborative decision-making suggestions and achieve data interoperability and sharing; Intelligent Service and Optimization Module: Used to build intelligent service models, automatically generate services based on service requests, track and provide feedback on service progress in real time, optimize and update models, and improve service response efficiency; The method for creating the data fusion and collaboration model includes: H1: Data Processing: Standardize and clean the collected multi-source data of the smart park, and perform coordinate transformation and timestamp synchronization according to a unified spatiotemporal benchmark to obtain the processed multi-source data of the smart park. H2: Feature Fusion: Based on the spatiotemporal-semantic association knowledge network, the processed multi-source data of the smart park are cross-modal feature associations to establish a fusion database; H3: Collaborative Decision Making: Based on big data, analyze and integrate data from the database, mine the relationships between data, and generate collaborative decision-making suggestions; The method for constructing the spatiotemporal-semantic association knowledge network includes: N1: Data Labeling: Assigning preliminary semantic labels to the processed multi-source data of the smart park; N2: Geospatial Association Construction: Based on the processed multi-source data of the smart park, construct a spatial geographic model of the smart park; calculate the spatial relationships between different spatial elements, including: distance, adjacency, and containment; N3: Spatiotemporal element extraction: Analyze and process multi-source data of the smart park, extract time patterns, mine time patterns, and obtain spatiotemporal elements; N4: Semantic association knowledge extraction and fusion: Extract semantic knowledge of smart parks from multi-source data and relevant literature on smart parks after processing, and construct a semantic knowledge base; based on the semantic knowledge base, associate the multi-source data and semantic knowledge of smart parks after processing. N5: Network Construction: Using the processed multi-source data of the smart park, spatiotemporal elements, semantic knowledge, and spatial geographic models and their spatial relationships as nodes, and the association relationships and spatial relationship weights between nodes as edges, a spatiotemporal-semantic association knowledge network is constructed. The method for creating the intelligent service model includes: S1: Service Function Integration: Integrating various service function modules; S2: Process Definition: Defines the processing flow for various services, including task allocation, processing time limits, and workflow paths; S3: Information Input: The user inputs a service request; S4: Service Execution: Based on the processing flow, determine the service type according to the service request, automatically generate a service, and transfer it to the relevant service; S5: Service Feedback and Optimization: Track the progress of service request processing in real time and provide feedback to users; after the service is completed, push a service evaluation survey to users, analyze the collected service evaluation data, and optimize the processing flow based on the analysis results.

2. The smart park collaborative service system based on cloud computing and big data according to claim 1, characterized in that: The steps for generating collaborative decision-making recommendations include: P1: Data Analysis: Perform multi-dimensional analysis on the data in the integrated database. The multi-dimensional analysis includes: time dimension, spatial dimension, and modal dimension, to obtain the data analysis results; P2: Association rule mining: Based on the data analysis results, mine the association relationships between data in the integrated database; and predict future data trends and events based on the association relationships to obtain prediction results; P3: Generate decision-making suggestions: Construct decision-making scenarios, determine decision indicators and influencing factors; based on correlations and prediction results, generate collaborative decision-making suggestions for the corresponding scenarios.

3. The smart park collaborative service system based on cloud computing and big data according to claim 1, characterized in that: The feature fusion method includes: M1: Cross-modal feature association: The processed smart park data is spliced ​​to generate a basic feature vector; multimodal features are weighted and aggregated to obtain a multimodal feature vector; M2: Dimensionality Reduction and Optimization: Dimensionality reduction is performed on the spatiotemporal-semantic association knowledge network and multimodal feature vectors to obtain dimensionality-reduced feature vectors. Feature weights are then adjusted to establish a fusion database.

4. The smart park collaborative service system based on cloud computing and big data according to claim 3, characterized in that: The method for cross-modal feature association includes: Q1: Concatenation: Connect the feature vectors of different modalities aligned within the same spatiotemporal unit to obtain the basic feature vector; Q2: Calculate feature weights: Automatically learn the specific content of the basic feature vectors, obtain the dynamic weight of each feature dimension, and highlight key information; Q3: Weighted aggregation: The calculated feature weights are summed with the basic feature vectors to obtain the multimodal feature vector.

5. A smart park collaborative service system based on cloud computing and big data according to claim 3, characterized in that: The method for adjusting feature weights includes: E1: Normalized dimensionality reduction features: The dimensionality-reduced feature vectors are standardized, and the initial weights are redistributed based on the contribution of the feature variance. E2: Calculate dynamic weights: Analyze the semantic relationships between multimodal features and calculate dynamic weight coefficients; E3: Similarity-driven adjustment: Calculate the similarity matrix between feature vectors and adjust the feature weights according to the strength of the correlation.

6. The smart park collaborative service system based on cloud computing and big data according to claim 1, characterized in that: The service execution steps include: R1: Determine Service Type: Parse the user's service request, identify the service type, and match it with a predefined processing flow; R2: Service Generation: Based on the processing flow, service tasks are assigned to the corresponding processing nodes, and processing progress data is synchronized to the fusion database in real time.

7. The smart park collaborative service system based on cloud computing and big data according to claim 1, characterized in that: The service feedback and optimization includes handling service timeouts and rule conflicts. The handling methods include: automatically triggering an early warning mechanism and transferring the call to a manual review node, notifying the user of the reasons for the service timeout and the solution, and updating the fusion database.

Images