Predicting Litecoin price movement in a pre-defined trading volume window using Random Forest model

Abstract

Over the past years, there has been a growing interest in cryptocurrency markets. In this context, price forecasting initiatives that aid in the decision-making process of investors and market participants have emerged and drawn the interest of academia and the financial technology industry. In this paper, we present a machine learning classification model that forecasts the price direction - top, modeled as 1, or neutral or bottom, modeled as 0 - of Litecoin (LTC) over the forecast horizon equivalent to volume-wise samples of 100 thousand LTC. For modeling, we adopt a random forest classifier, achieving an Area Under the Receiver Operating Characteristic curve (AUROC or AUC) score of 0.65 on the hold-out, out-of-time test subset.