Currency Price Prediction using Machine Learning Algorithms
A Kamini, Prince Sood · YMER Digital · 2022
Bitcoin, being the most well-known blockchain technology, has recently received a lot of attention in the fields of economics and finance. The purpose of this dissertation is to determine if new machine-learning models can forecast better than traditional models. This research compares the accuracy of bitcoin price prediction using two different models in terms of forecasting errors: Long-Short Term Memory (LSTM) against Auto Regressive Integrated Moving Average (ARIMA), and Python routines were used. The Federal Reserve Economic Statistics was used to compile Bitcoin price data from 2017-06-18 to 2019-08-07. To compare the results of both models, the data was separated into two subgroups: training (83.5%) and testing (83.5%). (16.5 percent). In most cases, the literature shows that LSTM outperforms ARIMA. According to RMSE and MAE, In this dissertation, LSTM forecasts of bitcoin prices outperform average ARIMA predictions by 92 percent and 94 percent, respectively