Hybrid Long Short Term Memory And Convolution Neural Network Model For Volatile Time Series Prediction

Abstract

In recent years significant breakthrough has occurred at the cross roads of artificial newlineintelligence using neural networks have arisen as effective tool for forecasting complex and newlinevolatile time series data. Deep learning neural network models have shown considerable newlineresults in detecting patterns and making predictions on large volumes of data. The Long short newlineterm memory neural networks and Convolution neural networks play a vital role in the deep newlinelearning process. In this research work the Long short term memory neural network, newlineConvolution neural network and long short-term memory fused convolution neural network newlineare developed for predicting of weather conditions using weather time series data. newlineFurther LSTM neural network with Aquila optimizer, CNN with Aquila optimizer and hybrid newlineLSTM, CNN, Support vector machine-based algorithm are developed for stock market newlineprediction using stock price time series data. The weather prediction data set is obtained from newlinezenodo.org database which involves European weather conditions for 18 locations. It contains newlinedaily meteorological data for these locations across various time periods. It include 3654 daily newlinerecords covering weather conditions for each of the 18 locations resulting into a large , newlinecomplex and volatile data set. The stock market time series data is obtained from financial newlinewebsites like yahoo finance and money control. The stock price time series data for 5major newlinecompanies are considered for our prediction process. newline