FORECASTING STOCK MARKET INDEX BASED ON PATTERN-DRIVEN LONG SHORT-TERM MEMORY

Abstract

Stock trend prediction is an important area of study for researchers and practitioners. In recent years, along with traditional statistical prediction models, machine learning and deep learning techniques have been increasingly adopted in various financial studies. Long Short-Term Memory (LSTM) is one of the deep learning models for predicting time -series data. In the case of vanilla LSTM, shared weights are learned based on all available data; hence, it is difficult to accurately learn patterns and predict the future value from a subset of data. In this paper, a pattern -driven hybrid model that combines an LSTM with an unsupervised learning algorithm is proposed for precise prediction of stock prices. The performance of the hybrid model is evaluated using Korea stock index data. The results demonstrate that the proposed model outperforms traditional recurrent neural network (RNN) and LSTM models.