ADS Capstone Chronicles Revised
1 Guardians of the Crypto: A Streamlit Application for Enhanced Price Prediction and Informed Decision-Making
Historical Data for Accurate Cryptocurrency Forecasting
Mirna Philip Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of San Diego mphilip@sandiego.edu
Justin Farnan Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of San Diego jfarnan@sandiego.edu
Arya Shahbazi Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of San Diego ashahbazi@sandiego.edu
refining technical indicators to improve predictive accuracy. Future work will focus on automating data extraction and forecasting processes to enhance operational efficiency and ensure timely updates. A Streamlit application was developed to showcase various trading indicators and forecasted percentage change predictions. By providing accurate and timely price predictions, “ Guardians of the Crypto ” aspires to empower users to navigate the cryptocurrency market more effectively, enabling informed trading and investment decisions. This application is poised to become an indispensable tool for novice and experienced traders, offering a comprehensive solution to manage risks and capitalize on opportunities in the dynamic cryptocurrency market. KEYWORDS Cryptocurrency Prediction, Data Science, Machine Learning, Time Series Analysis, Predictive Modeling, Financial Technology 1 Introduction The cryptocurrency market is highly unpredictable, with prices fluctuating significantly in short periods. This volatility makes it challenging for investors and traders to make well-informed decisions. This project,
ABSTRACT The “Guardians of the Crypto” project aims to address the challenges posed by the high volatility of the cryptocurrency market. It offers a comprehensive solution to manage risks and capitalize on opportunities. By developing an algorithm capable of predicting future price changes, this study employed advanced data science techniques and machine learning algorithms to analyze the historical price data of various cryptocurrencies, including Bitcoin (BTC), Ethereum (ETH), Litecoin (LTC), Cardano (ADA), etc. The methodology involved meticulous data cleaning, feature engineering, and preprocessing to ensure data integrity and enhance model accuracy. Although slightly overfitted, Ridge Regression emerged as the most consistent and reliable model, demonstrating superior performance in handling multicollinearity and providing unwavering predictions across all evaluated metrics, specifically the adjusted R-squared. Despite the inherent unpredictability of the cryptocurrency market, the model showcased robust forecasting capabilities for most cryptocurrencies, although some inconsistencies were noted with BTC USD. The study ’ s findings underscore the importance of using higher frequency data, integrating sentiment analysis, and continuously
68
Made with FlippingBook - Online Brochure Maker