Open AccessAnalytics of stock market prices based on machine learning algorithms
Author(s) -
Puteri Hasya Damia Abd Samad,
Sofianita Mutalib,
Shuzlina Abdul-Rahman
Publication year - 2019
Publication title -
indonesian journal of electrical engineering and computer science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.241
H-Index - 17
eISSN - 2502-4760
pISSN - 2502-4752
DOI - 10.11591/ijeecs.v16.i2.pp1050-1058
Subject(s) - support vector machine , random forest , computer science , stock market , algorithm , machine learning , artificial intelligence , analytics , profit (economics) , stock (firearms) , data mining , economics , engineering , mechanical engineering , paleontology , horse , biology , microeconomics
This study focuses on the use of machine learning algorithms to analyse financial news on stock market prices. Stock market prediction is a challenging task because the market is known to be very volatile and dynamic. Investors face these kinds of problems as they do not properly understand which stock product to subscribe or when to sell the product with an optimum profit. Analyzing the information individually or manually is a tedious task as many aspects have to be considered. Five different companies from Bursa Malaysia namely CIMB, Sime Darby, Axiata, Maybank and Petronas were chosen in this study. Two sets of experiments were performed based on different data types. The first experiment employs textual data involving 6368 articles, extracted from financial news that have been classified into positive or negative using Support Vector Machine (SVM) algorithm. Bags of words and bags of combination words are extracted as the features for the first experiment. The second experiment employs the numeric data type extracted from historical data involving 5321 records to predict whether the stock price is going up (positive) or down (negative) using Random Forest algorithm. The Rain Forest algorithm gives better accuracy in comparison with SVM algorithm with 99% and 68% accuracy respectively. The results demonstrate the complexities of the textual-based data and demand better feature extraction technique.