Using training set selection methods to improve text mining on market prediction via news headlines

摘要

Nowadays, with the development of technology and social media, people's opinions and news about a particular product or currency is significantly expanded. In addition to expanding volumes of text data, the unstructured characteristics of them makes the analysis of these types of data with the vital challenges. In this study, the proposed method utilized different Training Set Selection (TSS) approaches like: ModelCS, ENN, MultiEdit, AllKNN, POP and MENN. Moreover, the target's features have been selected to improve the prediction of currency market (dollars, euros) based on news headlines. The evaluation of the proposed method has been done by 3-layered algorithm and the experimental results reveal that the proposed method has achieved more robust accuracies than the others.