Continuous Queries on Streaming Data
Master thesis
Permanent lenke
http://hdl.handle.net/11250/2565274Utgivelsesdato
2018Metadata
Vis full innførselSamlinger
Sammendrag
We are all living in a world that is becoming more digital for every day. While people are connecting to the Internet, there has been an explosion of applications that are being used on a daily basis. All of these applications have especially one thing in common - they are all generating data that are valuable for stakeholders to understand their users. Despite that, few know how to take advantage of this information. To unlock the information hidden in these applications, the concept of visualizations of both stored and streaming data has been explored. Furthermore, this project has investigated how a system can be implemented that makes it possible to visualize relevant information based on the interests of users. To filter out relevant content, the system performs continuous queries defined by the user. The filtering allows the system to avoid unnecessary storage and indexing of irrelevant data, in addition to making the visualization both intuitive and straightforward to interpret as all information that would not provide the user with valuable insights are excluded. The project has also explored how Machine Learning can be used to unlock valuable information about the data, and thus a sentiment analysis was performed. This project has utilized available information provided by Twitter as proof of concept. The system has been implemented to be modular, allowing other sources of streaming data to be used. The proposed solution should also consider the various aspects of big, streaming data and tolerate high enough throughputs of events. Our experiments show the practicality and feasibility of the proposed approach.