A Hybrid Fruit Fly ELM Framework for Stock Index Price Movement Prediction

Predicting the upward or downward direction of financial time-series data is very challenging due to complex factors involved in the data such as volatility, non-linearity, and sudden variations. This task requires a reliable and robust prediction model which can not only predict the direction accurately but also be faster than traditional algorithms. Extreme learning machine (ELM) is among the leading algorithms for such task at hand due to its extremely fast training and superior adaptation to new data. This study proposes a hybrid framework integrating fruit fly optimization with extreme learning machine (FFO-ELM) to predict the forthcoming movements of stock indices. Additionally, the performance of the advocated FFO-ELM framework is compared with differential evolution-based ELM (DE-ELM), particle swarm optimization-based ELM (PSO-ELM) and singleton ELM. The empirical analysis has been done utilizing BSE SENSEX and NIFTY 50 stock indices and the outcomes indicated that the FFO-ELM model performed comparatively better in terms of accuracy, fl-score, and g-mean in contrast with other inspected models.