Predictive Analytics for Cryptocurrency Prices Using Machine Learning

Abstract

This paper explores the use of big data in decision-making within the digital economy, with a specific focus on predictive analytics for Bitcoin price. Machine learning models are used as they are crucial for predicting Bitcoin prices due to their ability to analyze complex, volatile market data and automate trading decisions. These models enhance risk management and market efficiency by adapting to new trends and performing real-time analysis   It compares various machine learning models including Support Vector Machine (SVM), Decision tree regression, Random Forest, Long Short-Term Memory (LSTM), ARIMA, Linear Regression, and Quadratic Discriminant Analysis (Quaddisc). Python with Scikit-learn module and Pandas package along with others are used to assess the predictive analysis of the dataset. The study finds that Support Vector Machine (SVM) outperforms other models with an accuracy of 87%, compared to 64.37% for quadratic discriminant analysis and 52.9% for Random Forest. These results underscore SVM's superior efficiency in predictive analytics of Bitcoin prices. The paper also emphasizes the importance of fostering a data-driven culture within organizations to leverage big data analytics effectively, enhancing decision-making processes through comparative studies of these models.