Optimisation-Enabled Transfer Learning Framework for Stock Market Prediction

Abstract

Stock market prediction is a vital task with high attention for gaining attractive profits with proper decisions to invest. Predicting the stock market is becoming a major challenge nowadays due to chaotic data, non-stationary data, and blaring data. Hence, it’s challenging for investors to invest money to make profits. Many techniques are developed to predict stock market trends, but each differs based on time and year. In this paper, hybridised optimisation algorithm, namely the proposed Gannet Ladybug Beetle Optimisation (GLBO), is used for training Transfer Learning (TL), which considers Convolutional Neural Network-enabled Long Short-Term Memory (CNN-enabled LSTM). This TL is responsible for predicting the stock market from augmented data. Here, the stock market dataset of two companies is used as input time series data. Moreover, many features are extracted from those input data and then from those features, necessary features are selected based on Motyka similarity. The bootstrap method is used in this paper for data augmentation. Also, GLBO is hybridised with Gannet Optimisation Algorithm (GOA) along Ladybug Beetle Optimisation (LBO) algorithm. Furthermore, the proposed model is verified for its performance capability based on three metrics, Mean Squared Error (MSE), Root Mean Squared Error (RMSE), as well as Mean Absolute Percentage Error (MAPE) with values of 5.8%, 24.1%, and 78.8%.