LOCUST SWARM OPTIMIZED LOGISTIC DECISION TREE BASED TECHNICAL INDICATOR CLASSIFIER FOR STOCK MARKET PREDICTION

Abstract

As the economy has heightened swiftly in current years, more and more people have started investing their money into the stock market. Owing to this stock market prediction is considered as a pivotal venture and one that has proven to be more advantageous than others. Investors countenance notable issues making stock market-associated predictions as a consequence of dearth of movement and presence of noise. This study presents a detailed investigation of the selection of a minimal number of relevant technical indicators with the objective of increasing sensitivity, specificity, reducing training time and improving accuracy using Locust Swarm BoostARoota Optimized Logistic Decision Tree-based Classifier (LSBO-LDTC). The LSBO-LDTC method is split into three stages, namely, preprocessing, feature selection and classification for stock market prediction. In the first stage, data obtained from Stock Market Data - Nifty 100 Stocks (1 min) data are preprocessed to obtain cleaned ones by applying Log First Difference transformation-based Preprocessing. Then, in the second stage the preprocessed stock data is passed though the Locust Swarm BoostARoota Optimized feature selection for selecting relevant and significant technical indicators to obtain computationally efficient technical indicators for further classification. Finally in the third stage decision regarding suitable day regarding for buying or selling is performed using Logistic Decision Tree-based Classification for stock prediction. Experimental analysis is carried out on the parameters such as specificity, sensitivity, accuracy and training time with respect to number of stock data. Experimental results indicate a high performance of the proposed method for searching a global optimum stock market prediction to achieve high return on investment in comparison with other well-known machine learning methods.