Real-Time Anomaly Detection in Cryptocurrency Markets Using Hybrid Machine Learning Models

Abstract

                   Cryptocurrency markets are fast-moving and influenced by a variety of factors, which often leads to sudden and unexpected changes in price and trading activity. These irregular movements, known as anomalies, can occur due to reasons such as large transactions, breaking news, or shifts in public opinion. Identifying these patterns at the right time is important for better decision-making and reducing financial risk. However, many existing methods rely mostly on historical numerical data and are not fully effective in handling the dynamic and real-time nature of these markets.

This study proposes a hybrid machine learning approach for detecting anomalies in cryptocurrency markets. The model combines outlier detection, time-based analysis, and sentiment evaluation from social media and news sources. By using both numerical and textual data, the system provides a broader understanding of market behaviour. The approach is designed to work in real time, improving detection accuracy while reducing false signals. The results suggest that combining multiple methods offers a more practical and reliable solution for analyzing rapidly changing financial data.

Keywords: Cryptocurrency, Anomaly Detection, Machine Learning, Sentiment Analysis, Real-Time Analysis.