EXPLORE PUBLICATIONS BY COUNTRIES

Bitcoin, Cryptocurrency, Machine-Learning, Prediction.

Earlier to the advent of computers and the internet. Transactions such as buying, selling, hiring, and cash transfer are performed physically, hand-to-hand and/or face-to-face using hard-printed currency also known as traditional means. The recent advances in internet and networking technologies have significantly refurbished and improved the methods and limitations of the traditional ways, through cryptocurrency or digital money especially in terms of cost, speed, and access. These technologies which bring people together irrespective of geographical location have fashioned a revolution in trading and transaction processing; online transaction processing and real-time processing. However, like every other pioneering development, this is not without resistance from stakeholders, whom have been using the traditional means for long; its validity and legitimacy have been seriously challenged. In this work, several models leveraged to forecast bitcoin price were Linear Regression (LR), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Decision Tree (DT), Elastic Net Regression (EN), Lasso Regression (Lasso) and Ridge regression (RR). The models’ accuracies were determined and evaluated using Mean Absolute Error (MAE), Mean Square Error (MSE), and R Square Error (R2). It revealed good performance except for SVM which falls in the negative even after fine-tuning and improved performance. The LR led in performance, then EN, Lasso, and RR. Decision Tree on the other hand present an encouraging and challenging result. Whereas the SVM model presents worst-case prediction accuracy of -22.38%. Therefore, the linear regression model has the best fit for bitcoin price prediction amongst the algorithms tested and evaluated. The study would reduce researchers throughput by presenting firsthand model for price prediction, support vector machine need to be further studied to unravel reasons for its undesirable performance.