RB-EV BASED NEWS EVENT-DRIVEN STOCK FLUCTUATIONS DETECTION IN PAKISTAN’S MARKET THROUGH DATA MINING

Abstract

Stock market   plays a vital role in the development of firms, textile and industries, because it directly influences the economy of a country. Investor feels hesitation for investment because he does not know about the current situation of the stock market. The impacts of the news are more crucial for the investor as he invests in the appropriate company on the correct time. This research helps the investor for better understanding of the market situation and ups downs of Pakistan Stock Exchange (PSX). First, we need to parse data from news sites such as Daily times, daily Pakistan as well as from PSX. Afterward, we extract the news events according to categories to check their impact on PSX. For this purpose, we created a two-week date sliding window and then checked the impact of these dates on the stock market by using mapping technique. We set a Threshold to decide the importance of news, which affect positively and negatively on stock market. Then we perform validation of positive and negative events to check the accuracy of each event and overall accuracy of that method. After identifying event-related news using text mining and information extraction, the study uses association analysis to link these events to shifts in stock prices. The events' potential effects on the Pakistan Stock Market are then assessed using a rule-based classification technique (RB-EV Algorithm). The results of this study show positive events accuracy on 100 indexes is 66.71 % or negative events accuracy on 100 indexes is 67.02 %. The purpose of this research is to help the investors for decision-making.

Keywords : Events Extraction, Web Data Extraction, Information Extraction, PSX, KSE, Date sliding window, Mapping technique, Threshold, Rule-Based classification.