A Study on Mining Financial News to Predict S&P 500 Index Trend

Abstract

The stock market is flooded with a bunch of financial news every day Economist Eugene Fama (1970) proposed Efficient Market Hypothesis (EMH) stated that stock prices in the Semistrong-form efficient market already fully reflect publicly available information Is it possible that investors can get exceed returns from the market via daily financial news? To answer this question our study use text mining to obtain a lot of financial news from the Bloomberg website we quantify financial news and turn them into financial sentiment index Then we utilize financial sentiment index to build the technical analysis model for predicting S&P 500 index trend; we conduct the Vector Autoregression(VAR) statistical test the result shows that there is no significant relation between financial sentiment index and S&P 500 index This study finds that the information of financial news does not affect the trend of the S&P 500 index until the intensity fluctuation of the information is strong enough We mine the textual data from financial news to gain the useful information through LM (Loughran & McDonald 2011) OL (Hu & Liu 2004) and MPQA (Wilson et al 2005) three lexicons moreover create the news sentiment index We consider the news sentiment index as an indicator of stock technical analysis If the fluctuation range of the news sentiment index does not exceed 2 5 standard deviations the S&P 500 index continuously will be moving in its original direction When the volatility of the news sentiment index exceeds 2 5 standard deviations it will have a significant impact on the S&P 500 index We employ the discovery to simulate trading strategy in S&P 500 index and the returns for LM OL MPQA three lexicons were 6 12% 6 02% 4 42% and respectively

Date of Award	2019
Original language	English
Supervisor	Lih-Chyun Shu (Supervisor)