Please use this identifier to cite or link to this item: https://www.um.edu.mt/library/oar/handle/123456789/35865
Title: NEuRiTIS : linking NEws Reports with TIme Series Data
Authors: Muscat, Mark
Keywords: Time-series analysis
Data mining
Twitter
Discourse analysis
Issue Date: 2018
Citation: Muscat, M. (2018). NEuRiTIS : linking NEws Reports with TIme Series Data (Bachelor's dissertation).
Abstract: In the financial world, time series plays a major role as the record keeping, technical analysis and historical model medium of any and all monitored financial assets. It is not uncommon for the price of an asset being recorded, such as that of a stock, a commodity or a currency, to break its predicted course, increasing or decreasing in value unexpectedly. When such an event occurs, people refer to news sources and similar time series patterns from the past for any clues as to what caused such an anomaly, and to what degree. Through the behaviour of such exceptional anomalies, we observe a close relationship between the time series data of a financial asset's price and the news being reported. NEuRiTIS expands on this aspect by investigating the relationship between interesting trends in a financial asset's price, in our case trends in the price of crude oil, and related news articles. This is achieved by first detecting and extracting anomalies from the time series datasets, and then linking them with news reports published in the same time range as the anomalies. Finally we label each linked article with a sentiment mirroring the trend of the anomalies with which it is linked, and investigate the relationship between the sentiment of the articles and the trends of the time series anomalies. Observing the results obtained, we find that there is indeed a correlation between the sentiment of a news article and an anomaly happening in a relatively short time from the article's date of publication. This correlation however, decreases gradually the further away an anomaly occurs from the publication date of the news article. Through this correlation, we bridge the gap between time series data and news articles textual data. The research conducted throughout this study can act as a base for other time series analysis and textual data mining systems. Not only so, but it can also be used as a model for time series prediction techniques.
Description: B.SC.ICT(HONS)ARTIFICIAL INTELLIGENCE
URI: https://www.um.edu.mt/library/oar//handle/123456789/35865
Appears in Collections:Dissertations - FacICT - 2018
Dissertations - FacICTAI - 2018

Files in This Item:
File Description SizeFormat 
18BSCIT010.pdf
  Restricted Access
1.21 MBAdobe PDFView/Open Request a copy


Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.