The Design And Implementation Of A Local Event Detection System Using Geo-tagged Twitter Data

Nowadays, people live in a dynamic urban space in which every minute,several planned and unplanned events are happening at different locations. Andpeople can be delayed or have time unexpectedly due to unforeseen events thathappened in their real life. The local events detection become a need for life. Withthe development of social networks and the popularity of smart phones, thegeo-tagged social data makes it possible for local events detection system to provideaccurate events information.This paper describes the design and implementation of a local event detectionsystem using geo-tagged Twitter data. Briefly, the system is able to detect localevents in certain geographical area using real time data from the social networkTwitter. In order to reach high detection accuracy, the system utilizes some datamining strategies. First, the system perform clustering on the source data to obtaincandidate events. Then classification process will be performed to filter the localevents according to the feature of events. The combination of the clustering methodand classification method would make the local events detection algorithm moreaccurate.Another emphasis of this paper is the design and implementation of theDBSCAN clustering algorithm and logistic regression classifier. The traditionalversions of these two algorithms are all performed on single machine. However, inorder to face large amount of unstructured data generated by social networks, thispaper proposes the design and implementation of the two algorithms based onMapReduce Framework to deal with social big data.This thesis describes the local event detection system development lifecycle:requirement analysis, design, implementation and testing. Moreover, the detaileddesign and implementation of key techniques such as local events detectionalgorithm, DBSCAN clustering algorithm and logistic regression classifier is alsogiven in this paper.