Multiphase text mining predictor for market analysis

There has been a lot of research on the application of data mining or knowledge discovery in financial market predictions. In those researches, various data mining techniques are applied to predict stock price trends, index values, currency exchange rates, volatilities etc. However, most of the existing studies in this area are based on numeric and structured data for example, historical price quotes, financial statements, interest and tax rates or with other quantifiable figures. Studies on mining textual and unstructured information like news, recommendation and comments from experts, postings from online forums and chat rooms, personal blogs and so on seems to be only an emerging area of study. The available textual data on the internet is usually of huge quantity and is much more informative than purely numeric information. This information helps people to identify the market behavior from a piece of financial news but also understand the reason why the market behaves this way. The purpose of this thesis is to explore this emerging research area and to give comprehensive experiments and comparisons on applications of various textual data mining techniques on financial market predictions.