Two-Stage Hybrid Feature Selection Approach Using Levy’s Flight Based Chicken Swarm Optimization for Stock Market Forecasting

Stock market forecasting is done by analyzing multivariate financial time series generated through technical analysis. However, high-dimensional data deteriorates the prediction performance due to irrelevant features that lead to higher computational costs. Feature selection is used to reduce data dimensionality and select the most informative features. A two-stage hybrid feature selection method is proposed to improve the performance of the forecasting model. In the first stage, a way to aggregate multiple filter methods is introduced as Multi-Filter Feature Selection (MFFS). Three filter methods are used for MFFS to scan the dataset from different aspects. In the second stage, Levy’s Flight-based Chicken Swarm Optimization (LFCSO) is proposed. Levy’s flight is introduced to update the position of Chickens to handle local optima and early convergence. The proposed MFFS reduces the computational cost by filtering the ambiguous features with reduced computational load for the second stage. Deep learning models are used for forecasting using a reduced feature set. Extensive experiments have been performed with three stock indices. The proposed model is assessed against the feature subsets obtained under different scenarios. Performance validation is done by comparing the proposed model and the existing work based on various performance metrics. The experimental result shows that the proposed model outperforms the existing models.