Classification of Stock Price Movement With Sentiment Analysis and Commodity Price: Case Study of Metals and Mining Sector

Abstract

The unstable nature and complex behavior of the stock market make the prediction or forecasting process very difficult. The high level of debt and the declining price-earning ratio have bad implications for investment in metals and mining sector. This paper proposes a classification model for stock price movement based on financial news data, historical stock prices and commodity price data. We experiment with Support Vector Machine (SVM), Naive Bayes, and K-Nearest Neighbor (KNN) algorithm. The classifier then categorized the price into “up”, “down”, and “constant”. The result shows that the best model is achieved by Naive Bayes Algorithm with an accuracy of 60% in three days period by combining copper price and sentiment analysis features.