Title:  Implementation of a Pattern Detection Framework for Spationtemporal Data Streams

Description:

With the rise of ubiquitous technologies such as mobile phone data and GPS, location-based  services (LBS) become more and more important in our everyday life. Heterogeneous spatiotemporal data offers new opportunities for location analytics systems. Due to public data  sets like the NYC taxi trip data and more progressive data analytics tools like Apache Flink we are able to discover mobility patterns in urban spaces to provide new location-based services. With the increasing demand for LBS and technological advances combined with  user-generated data these new services can provide detailed information for more precise predictions or recommendations. Movement paths and frequent pick-up or drop-off locations in  certian time intervalls can be identified by analyzing the taxi trip data. One major challenge is to evaluate spatiotemporal data like the taxi trip data because it contains different important aspects both in time and space. The new location-based services can  offer taxi drivers and passengers additional benefits. In addition, the LBS analyzes a constant  data stream of location data generated by mobile devices or other data sources. From this data  stream new events must be generated in real time to respond to the current location data. A  key problem in real time data processing is the detection of event patterns in data streams.  The main goal of this thesis is to design and implement a prototype of pattern detection framework that analyzes a constant spatiotemporal data stream. Corresponding events are generated based on that patterns. New events can trigger for example a new point of interest that is  nearby and interesting for the current user of that location-based service. For evaluation purposes a service can be implemented which provides advanced knowledge to cab drivers and  passengers based on events. With spatiotemporal data sets other proof of concept services have  been successfully implemented. The location analytics framework provided by the TU Berlin is capable of computing different  data aggregations with mutable metrics on the taxi data set. Mobile devices are utilized to create a data stream that constantly creates new location data. This data stream is processed by  the complex event processing engine in Apache Flink 3. For instance the engine can trigger a  new event when a new point of interest is nearby. Based on the current location of the passenger or taxi driver the LBS can respond with additional information regarding the current traffic  situation or points of interest.

Supervisor: Bersant Deva

Type:  Bachelor Thesis

Duration: 6 months