ABSTRACT
Time series forecasting is a big fuss or it can be said that it’s a trending topic that has many feasible implementations including stock prices prediction, weather prediction, trade design and capital allotment. Therefore, it contributes to a large number of applications in the controversial domain of blockchain. In the realm of blockchain, forecasting is necessary so as to predict the future prospects. Owing to its immutable and decentralised nature, blockchain facilitates the tracking of malicious activities via anomaly detection. In the recent past, machine learning is spreading its root in every domain and blockchain is not left unmarked by it. It enforces the system to learn from past experience, acknowledge the pattern and thus apply that knowledge in future. In this work, we measure the prophecy or predicting capability of the bitcoin time sequence. This work presents an extensive review of machine learning algorithms along with an implemented study of ARIMA model, LSTM model and XGBoost to predict the fluctuations on a monthly basis. Further, the paper introduces K-means clustering-based anomaly detection scheme that predicts anomalies on the basis of timestamps in Weighted price and Volume of bitcoin. It has been shown that prediction of the anomalies on the dataset yields favourable outcomes and surpasses the outcomes of forecasting in terms of accuracy. Finally, the paper concludes by highlighting several open research challenges in the field of study.
