An Optimized Neural Network based BitCoin Price Prediction

Abstract

<strong>The most accepted cryptographic money is the bitcoin, which is highly attracting the traders and investors for making buy or sell decisions. However, the prediction of the bitcoin prices is challenging due to its higher voltality. In this work, a new bitcoin prediction model is introduced with three major phases: Pre-processing, Feature Extraction and Prediction. The collected bit coin data corresponding to minute-by-minute and hour-by-hour data is subjected to pre-processing. From the pre-processed data, the original features are extracted along with the features based on technical indicators. Average True Range (ATR), Exponential Moving Average (EMA) and Relative Strength Index (RSI) are the technical indicators computed. All the extracted features are subjected to prediction phase, where the optimized Neural Network (NN) model is used. To make the prediction more accurate, the training of NN is carried out by the renowned Elephant Herding Optimization (EHO) via tuning the weight. Finally, the algorithmic analysis is carried out by varying the window size.</strong>