Public service platform system based on artificial intelligence

Abstract

The invention provides a public service platform system based on artificial intelligence, and aims to realize accurate service recommendation required by a user through a plurality of modules. The system comprises a user interaction module which is used for receiving input request data such as texts, voices and images of a user and processing the input request data in combination with time and position information. The data acquisition module is responsible for acquiring multi-source heterogeneous urban public service data. Based on the collected data, the knowledge graph construction module constructs a dynamic knowledge graph comprising public service entities, user entities and attributes thereof. And the intelligent analysis module performs time sequence analysis, spatial analysis and cross-domain association analysis by using the knowledge graph to generate service demand prediction and spatial demand distribution results. And the personalized recommendation module provides the combination of the target service and the associated service, which most accord with the request of the user, for the user through a demand matching unit, a position optimization unit and a service combination recommendation unit. According to the system, the response efficiency of public services can be improved, and personalized and multi-field service recommendation is provided for users.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a public service platform system based on artificial intelligence. Background Art

[0002] In the prior art, public service platform systems usually provide services through manual operations or basic data processing methods. These systems can collect users' basic information and service needs, and provide corresponding public services through preset rules. With the development of information technology, some systems have begun to use simpler algorithms for data analysis and processing to improve the efficiency and accuracy of services. However, most of these systems are based on data from a single source, and their response to user needs is limited to specific service areas, and they cannot effectively integrate information from different data sources.

[0003] However, the existing public service platforms face some limitations in responding to the increasingly complex urban service needs. First, users’ service requests are often diverse and dynamic, and it is difficult for existing systems to fully understand users’ complex needs. Second, the data collection of existing systems mostly relies on a single or fixed source, and it is difficult to integrate multi-source heterogeneous data, resulting in low accuracy of service recommendations. In addition, the system usually lacks cross-domain collaborative analysis when providing services, and fails to fully explore the correlation between different public services.

[0004] Therefore, it is necessary to develop a new public service platform system based on artificial intelligence. Summary of the invention

[0005] This application provides a public service platform system based on artificial intelligence to provide users with personalized, multi-field accurate service recommendations.

[0006] This application provides a public service platform system based on artificial intelligence, including:

[0007] A user interaction module, used to receive input request data from a user, wherein the input request data includes text, voice, image input by the user, and time information and location information when the user inputs;

[0008] A data collection module, connected to the user interaction module, for collecting multi-source heterogeneous urban public service data, wherein the urban public service data includes historical service record data, user demographic data, and urban events and policy change data;

[0009] A knowledge graph construction module is connected to the data collection module and is used to collect urban public service data and construct a public service knowledge graph, wherein the public service knowledge graph includes public service entities and their attributes, user entities and their attributes, time and geographic location information of public service entities and user entities, and the relationship and interaction history between public service entities and user entities;

[0010] An intelligent analysis module is connected to the knowledge graph construction module and is used to generate a service demand analysis result by combining the user's input request data and the constructed public service knowledge graph; wherein the intelligent analysis module includes:

[0011] A time series analysis unit is used to perform time series analysis based on historical data in the constructed public service knowledge graph and user input request data to generate service demand prediction results;

[0012] A spatial analysis unit, used to perform spatial analysis based on the geographic location information in the constructed public service knowledge graph and the time information input by the user, and generate spatial demand distribution results; and

[0013] A cross-domain association analysis unit is used to perform cross-domain association analysis based on the potential associations between different public service entities in the constructed public service knowledge graph, and generate a cross-domain service combination recommendation strategy;

[0014] A personalized recommendation module, connected to the intelligent analysis module, is used to use the results of the time series analysis, spatial analysis and cross-domain association analysis performed by the intelligent analysis module to provide accurate service recommendations for users; wherein the personalized recommendation module includes:

[0015] A demand matching unit, used to select a target service that best meets the user input request from the service demand prediction results generated by the time series analysis unit, and give priority to recommending the target service;

[0016] A location optimization unit, configured to preferentially recommend relevant services in or near the user's area based on the spatial demand distribution result generated by the spatial analysis unit and in combination with the user's geographical location information; and

[0017] The service combination recommendation unit is used to recommend multiple associated public service combinations to the user based on the service combination strategy generated by the cross-domain association analysis unit.

[0018] The technical solution provided by this application has the following beneficial effects:

[0019] (1) Improve label accuracy: Through the user interaction module combined with time information and geographic location information, the intelligent analysis module can perform time series analysis, spatial analysis and cross-domain association analysis based on the knowledge graph to provide users with accurate service recommendations that meet their personalized needs, significantly improving the pertinence of services and user experience. (2) The data collection module can collect historical service records, demographic data, urban events, policy changes and other information from multi-source heterogeneous data, ensuring that the system can analyze user needs from different dimensions and make effective responses, enhancing the comprehensiveness and reliability of data analysis. (3) Through time series analysis and spatial analysis, the intelligent analysis module can predict service demand at specific times and regions in the future, helping the system to optimize resource allocation and service provision methods in advance, avoiding resource waste, and improving the efficiency and response speed of public services. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of an artificial intelligence-based public service platform system provided in the first embodiment of the present application. DETAILED DESCRIPTION

[0021] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0022] The first embodiment of the present application provides a public service platform system based on artificial intelligence. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 The first embodiment of the present application provides a public service platform system based on artificial intelligence for detailed description.

[0023] The system includes a user interaction module 101, a data collection module 102, a knowledge graph construction module 103, an intelligent analysis module 104 and a personalized recommendation module 105.

[0024] The user interaction module 101 is used to receive input request data from a user, wherein the input request data includes text, voice, image input by the user, and time information and location information of the user when inputting.

[0025] Furthermore, the user interaction module is specifically used for:

[0026] Receive user input request data in various forms, including text input, voice input, and image input; recognize and analyze text and voice through natural language processing technology, and analyze pictures or scans uploaded by users through image recognition technology to generate structured request information;

[0027] When the user inputs a request, the user's time information and geographic location information are captured and recorded, and transmitted to the data processing module together with the input request data.

[0028] The user interaction module 101 is used to receive data in various forms input by the user, including text, voice, image and other input forms. In order to process input in different forms, the user interaction module 101 uses natural language processing (NLP) technology and image recognition technology. For text and voice input, the module first parses the input through natural language processing technology. The text input directly extracts key information such as the type, details, time, etc. of the service request through the parser. For voice input, the system first uses voice recognition technology to convert the voice into text, and then further parses the text through natural language processing to generate structured service request information. For image input, the user interaction module 101 processes the pictures or scanned copies uploaded by the user through image recognition technology. This technology can automatically identify key information in the image, such as forms, certificates, tickets, etc., and extract relevant content, converting unstructured image data into structured request information.

[0029] In addition, when the user inputs the request, the user interaction module 101 can also automatically capture and record the time information and geographic location information of the user input. The time information is usually automatically generated by the system timestamp, accurate to the second. The geographic location information is obtained through the GPS function of the user's device or the network location service. The user's time information and geographic location information are packaged together with the user's input request data and transmitted to the data processing module for subsequent data analysis and service recommendation. This information helps the system understand the user's service request background, such as when and where the request was made, so that the corresponding service can be matched more accurately.

[0030] Through the multimodal data processing of the user interaction module 101, the system can comprehensively process user requests in different forms such as text, voice, and image, and automatically capture the user's time and location during the input process to ensure the accuracy and completeness of the data. This enables the system to understand the user's needs in multiple dimensions and provide more personalized and accurate services.

[0031] The data collection module 102 is connected to the user interaction module and is used to collect multi-source heterogeneous urban public service data, wherein the urban public service data includes historical service record data, user demographic data, and urban events and policy change data.

[0032] Furthermore, the data acquisition module is specifically used for:

[0033] Collecting data related to city public services from multiple data sources, including historical service record databases, user demographic databases, government policies and city event systems;

[0034] Identify and process heterogeneous data from different data sources, including structured data, semi-structured data, and unstructured data;

[0035] Through data conversion and standardization steps, the collected data is converted into a unified format to ensure that all data can be effectively used by the system.

[0036] The data collection module 102 plays a vital role in the public service platform system based on artificial intelligence. It is responsible for collecting data related to urban public services from multiple different data sources. This module is connected to the user interaction module 101 and starts the data collection process by receiving user input data and system automatic demand recognition. The range of data sources is very wide, usually including but not limited to historical service record databases, user demographic databases, government policy update systems, and urban event management systems. In order to ensure that all dimensions of urban services can be effectively covered, the data collection module needs to have a strong multi-source heterogeneous data processing capability.

[0037] In the specific implementation, the data collection module first collects data from different data sources. The historical service record database provides long-term information on the use of public services, such as the types of services requested by citizens in the past period of time, the time of service completion, and service satisfaction. By analyzing these historical records, the system can understand the usage trends and demand distribution of different public services. The user demographic database provides demographic data such as the age, gender, occupation, and income level of urban residents. These data play a key role in accurately providing personalized public services. The government policy system and the city event management system provide the latest policy changes, emergencies and other background information related to public services. These data can reflect the dynamic changes of the urban environment in real time, thereby helping the system to make corresponding adjustments.

[0038] The data acquisition module also has the ability to process data in different formats, because data obtained from different sources are usually heterogeneous, that is, these data may exist in structured, semi-structured or unstructured forms. Structured data, for example, comes from traditional database systems, and the data has clear fields and formats. Semi-structured data, such as XML files or JSON format data, have a certain structure, but cannot be used directly like relational databases. Unstructured data may come from text, pictures, videos, etc. in social media or documents, and usually have no fixed format. The data acquisition module can effectively identify these different forms of data through built-in recognition algorithms and data conversion tools, and convert them into a unified format that the system can handle. This conversion step not only includes simple data format conversion, but also involves complex data parsing and standardization processes to ensure that all data can be used uniformly in the system.

[0039] In order to ensure the consistency and integrity of the data, the data collection module will standardize all collected data. Through the standardization step, the module integrates heterogeneous data into a unified format, removes redundant information, and ensures that all information can be effectively processed and analyzed by the system under the same framework. For example, historical service record data and user demographic data may use different time formats or geographic information representation methods. The data collection module will convert these inconsistent information into a unified time and location format to ensure that the content of different data sources can be effectively matched and associated in the subsequent intelligent analysis process. In addition, the data collection module will also ensure the integrity of the data. During the data collection process, if it is found that there is missing or incomplete information in some data sources, the module will automatically supplement the missing data through data completion algorithms or from other related data sources.

[0040] Finally, the data collection module transmits the converted and standardized data to the knowledge graph construction module 103, which constructs a knowledge graph containing user entities, public service entities and their relationships based on these data as the basis for intelligent analysis. Through such multi-level, multi-source heterogeneous data collection and processing, the system can dynamically analyze and intelligently recommend urban public service needs on a comprehensive and accurate basis.

[0041] The knowledge graph construction module 103 is connected to the data collection module and is used to collect urban public service data to construct a public service knowledge graph, wherein the public service knowledge graph includes public service entities and their attributes, user entities and their attributes, time and geographic location information of public service entities and user entities, and the relationship and interaction history between public service entities and user entities.

[0042] Furthermore, the knowledge graph construction module is specifically used for:

[0043] Based on the urban public service data collected from the data collection module, identifying the public service entity and the user entity;

[0044] Assign corresponding attributes to the identified public service entities and user entities, where the attributes of the public service entity include service type, service provider, service description, timeliness, and availability; the attributes of the user entity include demographic information, service preferences, and interaction history;

[0045] Based on the historical interaction data between the user entity and the public service entity, the relationship between the entities is constructed; wherein the relationship includes user request for service, user participation in service, user feedback, service provision timeliness, and service availability.

[0046] The main function of the knowledge graph construction module 103 is to construct a knowledge graph that can accurately reflect the relationship between public service entities and user entities based on the urban public service data obtained from the data collection module 102. The module first identifies two core entities in the system, namely public service entities and user entities. Public service entities refer to various services provided in the city, such as transportation, medical care, education, government services, etc. Each service is regarded as an entity; while user entities represent residents in the city or individuals who use these public services.

[0047] Once these entities are identified, the system assigns them their own attributes. For public service entities, the attributes that the knowledge graph building module can assign include service type (such as medical services, educational services, etc.), service provider (such as a specific government department or third-party agency), service description (specific description of the service content), timeliness (whether the service is valid within a specific time) and service availability (the current status of the service, whether it can be provided). The setting of these attributes enables the knowledge graph to comprehensively and clearly display the characteristics and functions of each public service entity.

[0048] For user entities, the knowledge graph construction module assigns attributes to each user based on demographic information, service usage preferences, and interaction history obtained from the data collection module. Demographic information includes the user's age, gender, occupation, place of residence, etc. Service preferences are information obtained through analysis of the user's historical service request records, helping the system understand what types of public services the user tends to use; and interaction history records all past interaction data between the user and the public service entity, such as when the user used which services, the completion status of the service, and the user's feedback.

[0049] Next, the knowledge graph construction module will build the relationship between the user entity and the public service entity based on the historical interaction data between the two. These relationships include not only simple records of user requests for services, but also the process of user participation in services, such as the start time, duration, and end status of a service. The module also records user feedback on the service, which can be based on user ratings, comments, or complaints about the quality of the service. The timeliness of service provision and the availability of services are also key relationship elements. This information can show whether the service can be provided on time when the user requests it, and whether the service is available or when it will be restored to an available state.

[0050] Through these processes, the knowledge graph construction module gradually builds a dynamic and updateable graph that shows how urban public services interact with users. This knowledge graph not only includes each entity and its attributes, but also shows the complex relationships between entities, such as which services are most popular in a specific time period, which users tend to use a certain type of service, and user satisfaction or feedback on the service. This enables the module to provide a solid data foundation for other modules in the system (such as the intelligent analysis module 104) to support subsequent time series analysis, spatial analysis, and cross-domain association analysis.

[0051] In general, the knowledge graph construction module 103 visualizes various public services in the city and their relationships with users in a structured and dynamic manner, ensuring that the system can grasp the use of public services in real time and provide detailed and accurate data support for subsequent intelligent analysis and precise service recommendations. The working principle of this module ensures the integrity and reliability of the knowledge graph, enabling the public service platform to effectively respond to complex urban service needs.

[0052] The intelligent analysis module 104 is connected to the knowledge graph construction module and is used to generate service demand analysis results by combining the user's input request data and the constructed public service knowledge graph; wherein the intelligent analysis module includes:

[0053] A time series analysis unit is used to perform time series analysis based on historical data in the constructed public service knowledge graph and user input request data to generate service demand prediction results;

[0054] A spatial analysis unit, used to perform spatial analysis based on the geographic location information in the constructed public service knowledge graph and the time information input by the user, and generate spatial demand distribution results; and

[0055] The cross-domain association analysis unit is used to perform cross-domain association analysis based on the potential associations between different public service entities in the constructed public service knowledge graph and generate cross-domain service combination recommendation strategies.

[0056] Furthermore, the timing analysis unit is specifically used for:

[0057] Based on the historical service data in the constructed public service knowledge graph, identify the time trend and cyclical changes of service demand;

[0058] Combine the user's input request data, including the user's service type requirements, request time, and geographic location information, match the user's current request data with the historical data in the knowledge graph, and identify the time distribution characteristics of similar service requests; by analyzing the historical performance of similar requests at different time points, predict the service demand change trend at the current time point and in the future time period;

[0059] By combining historical service trends and user request patterns, a service demand forecast result for a specific future time period is generated; the service demand forecast result includes the expected service demand, peak demand period and geographical location distribution.

[0060] Furthermore, the spatial analysis unit is specifically used for:

[0061] Based on the geographic location information in the constructed public service knowledge graph, identify the geographic distribution of different public service entities, including the location of public service providers, the coverage of service facilities, and the location of user service requests; match the historical geographic data recorded in the knowledge graph with the current user's geographic location information to identify the historical service demand pattern related to the current service request and geographic location;

[0062] Combined with the time information input by the user, the distribution of service demand in different geographical areas within a specific time period is identified; by analyzing the service demand in the same geographical location in a similar time period in historical data, a regional demand distribution pattern is generated to predict the hot spots of service demand in the future time period;

[0063] By combining the time information input by the user and the corresponding geographic location data, a spatial demand distribution result within the current time period is generated, wherein the spatial demand distribution result includes predicted service demand peak areas, demand concentrations and potential service shortage areas.

[0064] Furthermore, the cross-domain association analysis unit is specifically used for:

[0065] Based on the constructed public service knowledge graph, identify the potential relationships between different public service entities;

[0066] By analyzing the association between service entities in different public service areas, the rationality of service combinations is evaluated by combining the co-occurrence between service entities, user preferences, and interdependencies within and outside the areas.

[0067] The association strength of service entities is quantified, and a recommendation strategy for cross-domain service combinations is generated based on users' current requests, historical needs, and association relationships.

[0068] The core function of the intelligent analysis module 104 is to combine the user's input request data and the public service knowledge graph through a series of technical means to achieve time series analysis, spatial analysis and cross-domain association analysis to generate service demand prediction results and cross-domain service recommendation strategies. The following will explain in detail the implementation of each analysis unit from the perspective of technical means to ensure that those skilled in the art can understand and implement these technologies.

[0069] First, the technical implementation of the time series analysis unit is based on time series analysis models and machine learning algorithms. The unit relies on historical data extracted from the knowledge graph, which usually exists in the form of timestamps and marks the usage of each service entity at different time points. In order to identify the time trend and periodic changes of service demand, the time series analysis unit can use time series prediction models such as autoregressive integrated moving average model (ARIMA) or long short-term memory network (LSTM). These models can learn the changing laws of service demand from historical service records and automatically capture long-term trends, periodic fluctuations and short-term fluctuations. In addition, the time series analysis unit will also combine the user's input data, including service type, request time and geographic location information, and match the current user request with similar requests in the historical service data through similarity matching algorithms (such as K nearest neighbor algorithm (KNN)). After the matching is completed, based on the performance of historical data, the system can generate service demand forecast results within a specific time period. These forecast results include demand volume, peak demand period, etc., and are output to the personalized recommendation module for further service recommendations.

[0070] In order to improve the accuracy of the prediction, the time series analysis unit will also use cross-validation technology to verify and adjust the model to ensure that the model can provide effective predictions for different types of services. In addition, based on the prediction results of the model, the time series analysis unit will also make dynamic adjustments according to the actual service demand data to cope with possible nonlinear changes or emergencies.

[0071] Secondly, the spatial analysis unit uses geographic information system (GIS) and spatial statistical technology, combined with geographic location information extracted from the knowledge graph, to perform spatial demand distribution analysis. The core technology of this unit is the processing and analysis of spatial data, mainly including geocoding, spatial interpolation, and hotspot analysis. First, the spatial analysis unit needs to match the user's current geographic location information with the historical geographic data in the knowledge graph, which can be achieved through geocoding technology. Geocoding technology converts the user's geographic location (such as address or GPS coordinates) into a standardized geographic entity, so that it can be matched with historical service data in the knowledge graph.

[0072] After the matching is completed, the spatial analysis unit uses spatial interpolation algorithms (such as Kriging interpolation or inverse distance weighted interpolation) to predict the historical service demand of different regions. These interpolation methods can predict the service demand of areas that have not yet been collected based on known historical data. By performing spatial analysis on service requests in the same time period in the past, the system can generate a service demand distribution map for the current time point and a certain time period in the future, including dense areas of service demand and potential service shortage areas. In order to discover peak demand areas, the spatial analysis unit can also use hot spot analysis technology to identify locations where service demand is abnormally concentrated in a specific area, thereby providing the system with a basis for targeted resource allocation.

[0073] Finally, the cross-domain association analysis unit analyzes the potential correlation between different public service entities through the entity relationships in the knowledge graph and the data mining algorithm, and generates a cross-domain service combination recommendation strategy. The technical core of cross-domain association analysis is graph theory algorithm and collaborative filtering technology. The knowledge graph is essentially a complex relationship network, in which each service entity and user entity exists in the form of a node, and the relationship between nodes is connected by edges. In order to identify the potential correlation between different service entities, the cross-domain association analysis unit can use a graph embedding algorithm (such as DeepWalk or Node2Vec) to embed the service entity into the vector space, thereby quantifying the similarity and association strength between each entity. Through this embedding, the system can analyze the frequency of co-occurrence of service entities in different service fields and the similarity of user preferences, and identify cross-domain service combinations.

[0074] In addition, the cross-domain association analysis unit can also use collaborative filtering technology to generate a recommendation strategy for service combinations by analyzing the user's historical service usage behavior. Specifically, collaborative filtering technology can discover the collaborative relationship between different service entities by identifying the common behaviors of multiple users in different service fields. For example, the system can analyze which users use transportation services and medical services at the same time, and provide cross-domain service combination recommendations for other users based on these patterns.

[0075] In actual applications, the cross-domain association analysis unit will combine the association strength of service entities, the user's historical needs, preferences and current requests to generate a cross-domain service combination recommendation strategy. This strategy is not only based on the user's single needs, but also provides users with a more comprehensive, cross-domain service combination through association analysis to meet the user's diverse needs.

[0076] The entire intelligent analysis module 104 works together through time series analysis, spatial analysis and cross-domain correlation analysis, and finally transmits the analysis results to the personalized recommendation module to generate accurate service recommendations. These technical means ensure that the system can provide users with efficient and personalized public service solutions based on the complex relationship between time, geographical location and services.

[0077] The following is the reference implementation code of the intelligent analysis module.

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] The personalized recommendation module 105 is connected to the intelligent analysis module and is used to provide accurate service recommendations for users using the results of the time series analysis, spatial analysis and cross-domain association analysis performed by the intelligent analysis module; wherein the personalized recommendation module includes:

[0086] A demand matching unit, used to select a target service that best meets the user input request from the service demand prediction results generated by the time series analysis unit, and give priority to recommending the target service;

[0087] A location optimization unit, configured to preferentially recommend relevant services in or near the user's area based on the spatial demand distribution result generated by the spatial analysis unit and in combination with the user's geographical location information; and

[0088] The service combination recommendation unit is used to recommend multiple associated public service combinations to the user based on the service combination strategy generated by the cross-domain association analysis unit.

[0089] Furthermore, the demand matching unit is specifically used for:

[0090] identifying the service most relevant to the user input request from the service demand prediction results generated by the time series analysis unit;

[0091] The selected candidate services are prioritized according to the strength of the match, the target service that best meets the user's request is determined, and the target service is recommended to the user first.

[0092] Furthermore, the location optimization unit is specifically used for:

[0093] Get the user's current location P u (t), the matching service provider location P s (t), and obtain the historical performance data H of the service provider through the data collection module s ;

[0094] The recommendation score of the candidate service provider is calculated according to the following formula 1:

[0095]

[0096] Among them, f opt represents the recommendation score of the candidate service provider; d(P u (t),P s (t)) represents the user position P u (t) and the location P of the service provider s (t) The distance between wait represents the expected waiting time of the service provider; c(P u (t),P s (t)) represents the degree of traffic congestion, which is obtained based on real-time traffic data; w represents the user preference weight; H s represents the performance score of the service provider; α1, α2, α3 are weight parameters; ∈, γ, δ are smoothing parameters;

[0097] The calculated recommendation scores of the candidate service providers are sorted to obtain the service provider with the highest recommendation score, and the service provider with the highest recommendation score is displayed to the user.

[0098] Furthermore, the service combination recommendation unit is specifically used for:

[0099] According to the following formula 2, the optimization score SCO of user u for service combination S at time t is calculated:

[0100] SCO=ω1·C(S)+ω2·R(u,S,t)+ω3·E(u,S) (2)

[0101] Among them, ω1, ω2, ω3 are weight coefficients; C(S) is the internal synergy function; R(u,S,t) is the user correlation function; E(u,S) is the expected efficiency function;

[0102] The internal synergy function is calculated by the following formula 3:

[0103]

[0104] Where |S| is the number of services in the service combination S; sim(s i ,s j ) indicates service i and j The semantic similarity between the two services is calculated using the Word2Vec model. Word2Vec is a model based on word vector representation. It converts the description text of the service into vectors, and then calculates the cosine similarity between these vectors to obtain the semantic similarity between the services. The range of cosine similarity is [-1, 1]. The closer the value is to 1, the more similar the semantics of the two services are.

[0105] comp(s i ,s j ) indicates service i and j The complementarity score is obtained from historical data using a frequent itemset mining algorithm. This algorithm analyzes which service combinations users often use together. The frequent occurrence patterns of service combinations indicate that these services are complementary. The complementarity score measures how often two services appear together in real applications, thereby evaluating whether they can complement each other to meet user needs.

[0106] The user relevance function is calculated by the following formula 4:

[0107]

[0108] Among them, cos(V u,t ,V s,t ) represents the cosine similarity of the feature vectors of user u and service s at time t. User feature vector V u,t and the service feature vector V s,t It is constructed based on user behavior data and service attribute history data. The user's feature vector usually includes the user's preferences, historical usage patterns, etc., while the service's feature vector is based on the service's attributes and description information. The cosine similarity can be used to measure the match between the user and the service. The higher the cosine similarity value, the higher the preference match between the user and the service.

[0109] I u,sIndicates the number of historical interactions between user u and service s. This metric reflects the interaction history between the user and the service, and smoothes the impact of historical data on current decisions through the logarithm of the number of interactions. If a user has frequently used a service in the past, the relevance score of the service will increase.

[0110] The expected efficiency function is calculated by the following formula 5:

[0111]

[0112] Among them, T s,u It represents the expected benefit of user u from using service s. The expected benefit can be estimated based on the user's historical rating of the service or the actual utility of the service, reflecting the benefits obtained by the user from using the service.

[0113] D s Represents the expected time cost of service s. The expected time cost refers to the time required for a user to use a service. The system estimates the average service duration based on historical data.

[0114] D max The maximum total time cost acceptable to the user. This is usually a threshold set by the user, indicating the maximum time the user is willing to spend in the service combination. If the service time cost in the combination exceeds this threshold, the system will impose a corresponding penalty on the score of the combination.

[0115] H(S) represents the entropy of the service combination S.

[0116] In order to calculate the entropy of the service combination S, the Shannon entropy formula related to information theory can usually be used. The specific calculation process is as follows:

[0117] 1: First, it is necessary to classify each service in the service portfolio S. Each service can be divided according to its type, function, category and other characteristics. For example, services can be divided into transportation services, medical services, education services, etc.

[0118] 2: For each category in the service portfolio S, calculate the ratio of the number of services in this category to the number of all services in the portfolio. This can be expressed as:

[0119]

[0120] where p i is the probability of class i.

[0121] 3: For each category in the service combination S, use the Shannon entropy formula to calculate the entropy of the combination S:

[0122]

[0123] Among them, pi is the probability of the i-th type of service; log(p i ) is usually taken as the natural logarithm or the logarithm with base 2; n is the total number of service categories.

[0124] The optimization scores SCO of multiple service combinations S are compared, and the service combination with the highest score is selected and recommended to the user.

[0125] The core task of the personalized recommendation module 105 is to provide users with personalized public service recommendations that best meet their needs by integrating the results of time series analysis, spatial analysis, and cross-domain association analysis, combined with the user's specific needs, geographic location, and historical interaction data. This module is specifically composed of a demand matching unit, a location optimization unit, and a service combination recommendation unit, each of which is responsible for the service recommendation process in different dimensions.

[0126] The demand matching unit first uses the service demand prediction results generated by the timing analysis unit to filter out the services that are most relevant to the user's input request. This process relies on the time trend analysis results of historical data and the service request data of the current user. During the screening process, the system determines which services are most likely to meet the user's current needs based on the type of service, the user's needs, and the degree of matching of historical service requests. In order to achieve this match, the system uses similarity measurement technology to evaluate the match by calculating the similarity between the user's current request and historical requests (such as the cosine similarity between feature vectors). Based on the strength of the match, the demand matching unit prioritizes the filtered services, determines the target service that best meets the user's needs, and recommends this service first.

[0127] The location optimization unit uses the spatial demand distribution results generated by the spatial analysis unit and the user's current geographic location information to prioritize related services in or near the user's area. The technical implementation of location optimization is based on formula 1, which involves the calculation of multiple parameters, such as the user's current location P u (t) and the location P of the service provider s (t) The distance d(P u (t),P s (t)), and the traffic congestion level c(P u (t),P s (t)). In addition, the system also considers the waiting time T of the service provider wait and performance score H s, these data are dynamically acquired by the data collection module 102. Based on formula 1, the system calculates the recommendation score of each candidate service provider, sorts them by score, and recommends the service provider with the highest score to the user first. This process not only takes into account the geographical location, but also the quality of service, waiting time and traffic conditions, to ensure that users can get a convenient and efficient service experience.

[0128] The service combination recommendation unit further expands the dimension of recommendation, and provides users with cross-domain service combination recommendations through the service combination strategy generated by the cross-domain association analysis unit. This unit uses Formula 2 to calculate the user's optimization score SCO for the service combination S at time t, which includes three core indicators: internal synergy, user relevance, and expected efficiency. The internal synergy function C(S) is calculated based on the semantic similarity and complementarity scores between services. The semantic similarity uses the Word2Vec model to represent the word vectors of the service description text and calculate the cosine similarity; the complementarity score uses the frequent item set mining algorithm to identify which service combinations have complementary characteristics from historical data and calculate based on the frequent occurrence pattern of the combination. The user relevance function R(u,S,t) is measured by calculating the cosine similarity of the feature vectors of the user and the service. The feature vector is constructed by the user's behavior data and the historical data of the service attributes, and the number of interactions I u,s It is also included in the calculation to reflect the historical interaction intensity between the user and the service. The expected efficiency function E(u,S) evaluates the overall efficiency of the user's use of the service combination, through the user's expected benefit T s,u and expected time cost D s Calculations are performed and the complexity of the service combination is taken into account (represented by the entropy of the combination H(S)). Through these technical means, the system scores multiple service combinations and finally selects the combination with the highest score and recommends it to the user.

[0129] The entire personalized recommendation process is multi-dimensional, taking into account not only the user's needs and geographic location, but also the efficiency of service provision and the synergy of the combination. The personalized recommendation module can not only recommend a single service to users, but also provide users with a more comprehensive and integrated service experience through cross-domain service combination recommendations. This recommendation model ensures the intelligence and flexibility of the system, and can make dynamic adjustments based on changes in user needs and real-time data, thereby providing users with the best public services.

[0130] The following is a reference implementation code of the personalized recommendation module 105 .

[0131]

[0132]

[0133]

[0134]

[0135]

[0136]

[0137] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

Claims

1. A public service platform system based on artificial intelligence, characterized in that: include: A user interaction module, used to receive input request data from a user, wherein the input request data includes text, voice, image input by the user, and time information and location information when the user inputs; A data collection module, connected to the user interaction module, for collecting multi-source heterogeneous urban public service data, wherein the urban public service data includes historical service record data, user demographic data, and urban events and policy change data; A knowledge graph construction module is connected to the data collection module and is used to collect urban public service data and construct a public service knowledge graph, wherein the public service knowledge graph includes public service entities and their attributes, user entities and their attributes, time and geographic location information of public service entities and user entities, and the relationship and interaction history between public service entities and user entities; An intelligent analysis module is connected to the knowledge graph construction module and is used to generate a service demand analysis result by combining the user's input request data and the constructed public service knowledge graph; wherein the intelligent analysis module includes: A time series analysis unit is used to perform time series analysis based on historical data in the constructed public service knowledge graph and user input request data to generate service demand prediction results; A spatial analysis unit, used to perform spatial analysis based on the geographic location information in the constructed public service knowledge graph and the time information input by the user, and generate spatial demand distribution results; and A cross-domain association analysis unit is used to perform cross-domain association analysis based on the potential associations between different public service entities in the constructed public service knowledge graph, and generate a cross-domain service combination recommendation strategy; A personalized recommendation module, connected to the intelligent analysis module, is used to use the results of the time series analysis, spatial analysis and cross-domain association analysis performed by the intelligent analysis module to provide accurate service recommendations for users; wherein the personalized recommendation module includes: A demand matching unit, used to select a target service that best meets the user input request from the service demand prediction results generated by the time series analysis unit, and give priority to recommending the target service; A location optimization unit, configured to preferentially recommend relevant services in or near the user's area based on the spatial demand distribution result generated by the spatial analysis unit and in combination with the user's geographical location information; and The service combination recommendation unit is used to recommend multiple associated public service combinations to the user based on the service combination strategy generated by the cross-domain association analysis unit.

2. The public service platform system based on artificial intelligence according to claim 1 is characterized in that: The user interaction module is specifically used for: Receive user input request data in various forms, including text input, voice input, and image input; recognize and analyze text and voice through natural language processing technology, and analyze pictures or scans uploaded by users through image recognition technology to generate structured request information; When the user inputs a request, the user's time information and geographic location information are captured and recorded, and transmitted to the data processing module together with the input request data.

3. The public service platform system based on artificial intelligence according to claim 1 is characterized in that: The data acquisition module is specifically used for: Collecting data related to city public services from multiple data sources, including historical service record databases, user demographic databases, government policies and city event systems; Identify and process heterogeneous data from different data sources, including structured data, semi-structured data, and unstructured data; Through data conversion and standardization steps, the collected data is converted into a unified format to ensure that all data can be effectively used by the system.

4. The public service platform system based on artificial intelligence according to claim 1 is characterized in that: The knowledge graph construction module is specifically used for: Based on the urban public service data collected from the data collection module, identifying the public service entity and the user entity; Assign corresponding attributes to the identified public service entities and user entities, where the attributes of the public service entity include service type, service provider, service description, timeliness, and availability; the attributes of the user entity include demographic information, service preferences, and interaction history; Based on the historical interaction data between the user entity and the public service entity, the relationship between the entities is constructed; wherein the relationship includes user request for service, user participation in service, user feedback, service provision timeliness, and service availability.

5. The public service platform system based on artificial intelligence according to claim 1 is characterized in that: The timing analysis unit is specifically used for: Based on the historical service data in the constructed public service knowledge graph, identify the time trend and cyclical changes of service demand; Combine the user's input request data, including the user's service type requirements, request time, and geographic location information, match the user's current request data with the historical data in the knowledge graph, and identify the time distribution characteristics of similar service requests; By analyzing the historical performance of similar requests at different time points, we can predict the changing trend of service demand at the current time point and in the future time period. By combining historical service trends and user request patterns, a service demand forecast result for a specific future time period is generated; the service demand forecast result includes the expected service demand, peak demand period and geographical location distribution.

6. The public service platform system based on artificial intelligence according to claim 1 is characterized in that: The spatial analysis unit is specifically used for: Based on the geographic location information in the constructed public service knowledge graph, identify the geographic distribution of different public service entities, including the location of public service providers, the coverage of service facilities, and the location of user service requests; match the historical geographic data recorded in the knowledge graph with the current user's geographic location information to identify the historical service demand pattern related to the current service request and geographic location; Combined with the time information input by the user, the distribution of service demand in different geographical areas within a specific time period is identified; by analyzing the service demand in the same geographical location in a similar time period in historical data, a regional demand distribution pattern is generated to predict the hot spots of service demand in the future time period; By combining the time information input by the user and the corresponding geographic location data, a spatial demand distribution result within the current time period is generated, wherein the spatial demand distribution result includes predicted service demand peak areas, demand concentrations and potential service shortage areas.

7. The public service platform system based on artificial intelligence according to claim 1 is characterized in that: The cross-domain association analysis unit is specifically used for: Based on the constructed public service knowledge graph, identify the potential relationships between different public service entities; By analyzing the association between service entities in different public service areas, the rationality of service combinations is evaluated by combining the co-occurrence between service entities, user preferences, and interdependencies within and outside the areas. The association strength of service entities is quantified, and a recommendation strategy for cross-domain service combinations is generated based on users' current requests, historical needs, and association relationships.

8. The public service platform system based on artificial intelligence according to claim 1, characterized in that: The demand matching unit is specifically used for: identifying the service most relevant to the user input request from the service demand prediction results generated by the time series analysis unit; The selected candidate services are prioritized according to the strength of the match, the target service that best meets the user's request is determined, and the target service is recommended to the user first.

9. The public service platform system based on artificial intelligence according to claim 1, characterized in that: The location optimization unit is specifically used for: Get the user's current location P u (t), the matching service provider location P s (t), and obtain the historical performance data H of the service provider through the data collection module s ; The recommendation score of the candidate service provider is calculated according to the following formula 1: Among them, f opt represents the recommendation score of the candidate service provider; d(P u (t),P s (t)) represents the user position P u (t) and the location P of the service provider s (t) The distance between wait represents the expected waiting time of the service provider; c(P u (t),P s (t)) represents the degree of traffic congestion, which is obtained based on real-time traffic data; w represents the user preference weight; H s represents the performance score of the service provider, which can be obtained through user satisfaction feedback; α1, α2, α3 are weight parameters; ∈, γ, δ are smoothing parameters; The calculated recommendation scores of the candidate service providers are sorted to obtain the service provider with the highest recommendation score, and the service provider with the highest recommendation score is displayed to the user.

10. The public service platform system based on artificial intelligence according to claim 1, characterized in that: The service combination recommendation unit is specifically used for: According to the following formula 2, the optimization score SCO of user u for service combination S at time t is calculated: SCO=ω1·C(S)+ω2·R(u,S,t)+ω3·E(u,S)(2) Among them, ω1, ω2, ω3 are weight coefficients; C(S) is the internal synergy function; R(u,S,t) is the user correlation function; E(u,S) is the expected efficiency function; The internal synergy function is calculated by the following formula 3: Among them, sim(s i ,s j ) indicates service i and j The semantic similarity of the service description text is calculated by the Word2Vec model based on the word vector representation of the service description text, and the cosine similarity between the services is used as the service s i and j The semantic similarity of i ,s j ) indicates service i and j The complementarity score is obtained by identifying complementary relationships from historical usage data through frequent item set mining algorithm and calculating based on the frequent occurrence patterns of service combinations; The user relevance function is calculated by the following formula 4: Among them, cos(V u,t ,V s,t ) represents the cosine similarity of the feature vectors of user u and service s at time t, which is calculated based on the feature vectors of the user and the service. The feature vectors are constructed based on the historical data of user behavior and service attributes; I u,s Indicates the number of historical interactions between user u and service s; The expected efficiency function is calculated by the following formula 5: Among them, T s,u represents the expected benefit obtained by user u from using service s; D s represents the expected time cost of service s; D max is the maximum total time cost acceptable to the user; H(S) represents the entropy of the service combination S; The optimization scores SCO of multiple service combinations S are compared, and the service combination with the highest score is selected and recommended to the user.

Images