Comparative analysis of Vanilla LSTM and Peephole LSTM for stock market price prediction

Stock market price prediction is one of the most challenging problems in the field of time series forecasting because of the chaotic and volatile nature of stock market data. The currently existing technology is evolving in terms of algorithmic sophistication to provide better ways to predict behavior in highly volatile and noisy environments. A new trend in the machine learning and pattern recognition fields suggests that time-series prediction should be done using a deep nonlinear topology. Many papers on stock market price prediction have analyzed various machine learning algorithms like Artificial Neural Networks, ARIMA, CNN, RNN, SVM classifier, Fuzzy classifier, Random forest, and Bayesian model. Most suggest that a variant of RNN called LSTM outperforms the other algorithms. However, there is still a lesser-explored area within LSTM variants that can further be explored. This paper compares the performance of Vanilla LSTM with the other LSTM variant called Peephole LSTM. The experiments revealed that Peephole LSTM outperforms Vanilla LSTM by a significant margin.