Cryptocurrency-Price-Prediction-using-LSTM-based-deep-learnin

The cryptocurrency market continues to grow larger despite fluctuation in prices and there are currently over 18,000 cryptocurrencies in use with a combined market cap of £1.55 trillion. Investments in cryptocurrency have a high risk but offer great returns. Many people have started investing in cryptocurrencies to diversify their portfolio. There is a need for an accurate and efficient method of forecasting cryptocurrency prices so investors can make better decisions and make more profit. This research aims to develop a model for predicting the price of Bitcoin, Ethereum, Monero and Ripple using deep learning and evaluate its performance using certain metrics. This research should also be able to identify the optimal values of parameters like the number of epochs and dropout rates to obtain the best results. The proposed model uses a neural network with three LSTM (Long Short Term Memory), three dropout layers and one dense layer. This network is trained using historical time-series data. The actual and predicted prices of the selected cryptocurrency are plotted using python libraries. The proposed model is successful in predicting the prices and it works the best for predicting Ethereum prices with a normalized RMSE of 0.0564. The best performance is achieved with a dropout rate of 0.2 and 25 epochs. All the primary objectives of this research were achieved. Even though the model can predict most of the fluctuation in prices, it should not be used alone to make investment decisions as people may lose money. However, it can help to study cryptocurrency price trends.