Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.742021
Title: Stock Market Random Forest-Text Mining (SMRF-TM) approach to analyse critical indicators of stock market movements
Author: Elagamy, Mazen Nabil
ISNI:       0000 0004 7225 9388
Awarding Body: Staffordshire University
Current Institution: Staffordshire University
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
The Stock Market is a significant sector of a country’s economy and has a crucial role in the growth of commerce and industry. Hence, discovering efficient ways to analyse and visualise stock market data is considered a significant issue in modern finance. The use of data mining techniques to predict stock market movements has been extensively studied using historical market prices but such approaches are constrained to make assessments within the scope of existing information, and thus they are not able to model any random behaviour of the stock market or identify the causes behind events. One area of limited success in stock market prediction comes from textual data, which is a rich source of information. Analysing textual data related to the Stock Market may provide better understanding of random behaviours of the market. Text Mining combined with the Random Forest algorithm offers a novel approach to the study of critical indicators, which contribute to the prediction of stock market abnormal movements. In this thesis, a Stock Market Random Forest-Text Mining system (SMRF-TM) is developed and is used to mine the critical indicators related to the 2009 Dubai stock market debt standstill. Random forest and expectation maximisation are applied to classify the extracted features into a set of meaningful and semantic classes, thus extending current approaches from three to eight classes: critical down, down, neutral, up, critical up, economic, social and political. The study demonstrates that Random Forest has outperformed other classifiers and has achieved the best accuracy in classifying the bigram features extracted from the corpus.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.742021  DOI: Not available
Share: