Deep Learning with Recursive Neural Network for Temporal Logic Implementation

Abstract

Stock prediction in financial market is one of the exciting applications of deep learning (DL).Stock values fluctuate in accordance with time and hence suitability of Recursive Neural Network (RNN) as one of the model for prediction of stocks is a niche choice as a predictor.The temporal pattern of financial market is investigated with plain RNN and also with Long Short Term Memory (LSTM).In contrast to the classical time series forecasting, the widespread use of deep learning network and algorithms using temporal statistical relations built upon RNN have been finding increasing applications in time series domain analysis and provide a better yield in terms of performance.We have implemented temporal logic using Keras framework and show the results on the sequential data sets of stock prediction market.Our main focus is solely on stock prediction using time series forecasting, however, it can be extended to risk assessment, portfolio management etc.The imperative is to use this as a basis model for further investigation and explore additionally with trend forecasting, crypto currency forecasting and many more.Although this area is matured enough, with machine learning techniques, it provides tremendous opportunities for further investigations and improvements thereon.