Crypto Currency Price Prediction & Market Analysis

Authors:

Mirtipati Satish Kumar, S. Sunanda, P. Sai Karthik, B. Sravan Kumar, P. Jaswanth Sai Ram

Department of Information Engineering and Computational Technology,

MVGR college of Engineering (A), Vizianagaram 535005, Andhra Pradesh

Abstract - Cryptocurrency markets exhibit extreme price volatility shaped by a confluence of factors including technical market signals, investor sentiment, macroeconomic conditions, and evolving regulatory frameworks. Classical statistical approaches often fail to adequately model the intricate temporal dependencies inherent in digital asset pricing. This paper presents an integrated forecasting framework that combines deep learning with sentiment analysis to address these challenges in cryptocurrency price prediction and market trend analysis. The core predictive component employs a four-layer Bidirectional Long Short-Term Memory (Bi-LSTM) architecture augmented with an attention mechanism, trained on five years of historical price data. Sentiment signals are extracted from real-time cryptocurrency news feeds using the VADER lexicon-based analyzer. The feature space encompasses over 75 custom-engineered technical indicators spanning trend, momentum, volatility, volume, and support-resistance dimensions. Built on TensorFlow and Keras, the framework is evaluated across ten major cryptocurrencies—BTC, ETH, SOL, BNB, DOGE, XRP, ADA, AVAX, DOT, and LINK. Experimental results demonstrate 24-hour forecast accuracy with a Mean Absolute Percentage Error (MAPE) ranging from 2.5% to 4.2%, alongside a directional accuracy of 55% to 63%, outperforming conventional benchmark models. Predictions and sentiment insights are delivered through an interactive web-based dashboard enabling real-time monitoring and informed investment decision-making.

Keywords: Cryptocurrency, Price Prediction, Machine Learning, Deep Learning, Bi-LSTM, Sentiment Analysis, Gen-AI, Financial Forecasting.