Intelligent Model Building Of Social Media-based Spatiotemporal Analysis Workflow In Typical Scenarios

In recent years, along with the application and popularization of Internet, network communications and other technologies, social media applications, such as Facebook, Twitter, Sina Weibo, has attracted a large number of users and revolutionarily impacted the daily life on various aspects. Hundreds of millions of netizens use social media services every day to share information, communicate with friends, and express their opinions. With the popularization of the online map services, location-based services, and GPS enabled mobile devices, a growing number of social media platforms now allow user to report their locations when they post information. The ability of reporting locations not only brings more convenience to the social media users, but also made the social media data suitable for spatiotemporal analysis. The social media data is large in volume and its volume is rapidly growing for the huge user number of and the high frequency of their interactions. The social media data combines multiple dimensions together, including the social relationship, time, locations, contents, and so on. These characteristics make social media data a typical big data. The social media data with spatiotemporal connections cover almost all areas that human resident, update every seconds, and reflect the change of world from a huge group of people, which is not possible for the traditional geographic data and survey data. Researchers in the varies domains, such as information science, geography, and sociology, have noticed the value of social media data and apply spatiotemporal analysis to them in order to solve some application problems. These research findings have shown the great practical potential and broad prospects of social media-based spatiotemporal analysis.In the process of turning social media spatiotemporal analysis research into practical applications, the building of spatiotemporal analysis workflow is a bottleneck. It is difficult for an ordinary application user to quickly and accurately build a social media-based spatiotemporal analysis workflow. First, social media data have diverse sources, huge amount, and complex content. Different social media platforms may involve different user groups, different type of contents, different focus of attention, and different ways for data retrieving. How to find the right data from social media for a certain application need is a problem. Second, the analysis of social media data involves complex workflow and a large number of processing tools in diverse domains, including spatial analysis, time series analysis, statistical analysis, network analysis, and so on. How to choose available and proper processing tools to build a workflow that fulfills the application requirements is another problem. At present, the building of spatiotemporal analysis workflows still relies on the manual work of domain experts. In the literatures that use social media data for application research, some workflows for spatiotemporal analysis are constructed. However, these workflows have limited reusability. Modifications are required when you apply these workflows in your own applications. In one word, the current approaches for constructing social media-based spatiotemporal analysis workflow are far unable to meet the rapidly growing and diversifying application needs. How to effectively and formally represent the knowledges involved in the building of workflows in the existing researches, and how to use these knowledges for rapid and intelligent building of spatiotemporal analysis workflows in other circumstances are the key problems that must solve in the way to fulfill the application needs of social media-based analysis.To solve these problems, a bundled approach is proposed in this thesis. A knowledge description framework is proposed for modeling the knowledges involved in the social media-based spatiotemporal analysis workflows. And a method for intelligent building of spatiotemporal analysis workflows is proposed based on the semantic web technology and the AI planning technology. The proposed knowledge description framework provides the basic concepts in the social media domain, the way to model the semantics of social media data, and the way to describe, share and connect the workflow knowledges. The formally described knowledges can be recognized and reused by computers. The proposed method for intelligent building of spatiotemporal analysis workflows uses the formally described knowledge to recognize the applications requirements and builds workflows that adapt the scenarios. The main research contributions of this thesis are as follows:(1) The modeling of domain knowledge in social media-based spatiotemporal analysis are researched in this thesis. The semantic description method of social media concepts, datasets, analysis processes are proposed, which provides the domain knowledge basis for the intelligent workflow building. (2) The modeling of workflow knowledge in spatiotemporal analysis are researched. The semantic description method for the application requirements, application scenarios, and solutions for requirements is proposed, which make the workflows knowledge connected and reusable; (3) Based on the AI planning algorithm, the automatic building methods of social media-based spatiotemporal analysis workflow is proposed, which realized the requirement-oriented and scenario-aware building of workflows.In order to verify the feasibility and efficiency of this approach, two typical scenes, the emergency response to disasters, and the urban functional zoning, are selected in this thesis and domain knowledgebase and workflow knowledgebase are constructed correspondingly. The prototype for experiments is designed and implemented. Several social media-based analysis requirements are submitted to the prototype for workflow building tests. The results have shown that the approach proposed in this thesis can effectively reduce the complexity of the workflow building process, improve the quality of workflows and the efficiency of the building process, and speed up the converting of research findings in social media-related domains into practical applications.