The predictions of a daily stock price direction from the Thai news content by using natural language processing

Abstract

Factors affecting a stock price in the Stock Exchange of Thailand have several factors, including the various news. Hence, the concept of the daily stock price forecasting from the Thai news content using the natural language processing is studied so investors are able to forecast the stock price trends before the Stock Exchange of Thailand operates.  We made a study of the Thai news content with the tokenizer in Python version 3.7.1 from library pythainlp and then classification model was used for finding the most accurate values of the model and the tokenizer. This study was carried out the forecast of stock price trends in three days: 5th, 6th, and 7th February, 2020. One stock randomly chosen used simple random sampling from the following five stock groups: the ICT group, the ENERG group, the HELTH group, the COMM group, and the BANK group. The results revealed that the stock of Intouch Holdings Company (INTUCH randomly chosen by the ICT group is an efficient Gradient Boosting Classifier model when it is compared with forecasting and actual values of 100 %, the stocks of Thai Oil Public Company Limited (TOP) and Bumrungrad Hospital (BH), randomly chosen by the ENERG group and the HELTH group respectively, are not able to give us efficient models when they are compared with forecasting and actual values of 66.67 %. In addition, the stocks of CP ALL public company limited of COMM group and the stock of Kasikornbank Public Company Limited of the BANK group are efficient KNeighbors Classifier models when they are compared with forecasting and actual values of 66.67 %.