A study on the prediction of cryptocurrency price trend using k-means clustering and KNN classification

Abstract

The use of cryptocurrencies has increased all over the world. It is very hard to predict the future value of cryptocurrencies due to its high volatility. Therefore, cryptocurrency price prediction is a subject of interest to various individuals including investors. Numerous research has been done, however mostly it focuses on the use of supervised deep learning models, which are typically prone to overfitting and hence end up compromising the accuracy of the model. On the other hand, literature has given less attention to unsupervised approaches. This dissertation will investigate the predictability of cryptocurrency movement using an unsupervised learning technique, K-means clustering and a supervised learning technique, KNN classification. These algorithms will be compared with other deep learning models implemented in previous studies, to test if these less complex structures can achieve similar accuracies to more complex deep learning models. The predictive accuracy will be calculated and compared with the accuracy achieved by the state-of-the-art models proposed by Vella Critien et al in their paper titled 鈥淏itcoin Price Change and Trend Prediction Through Twitter Sentiment and Data Volume鈥� [9].