Genelleştirilmiş Toplamsal Modeller ile Bıtcoin için Yön Analizi

Abstract

Global economic uncertainties and rising inflation concerns have challenged the central role of traditional banking systems, limiting users' financial freedom. The emergence of Bitcoin and other cryptocurrencies has revolutionized the financial world. Bitcoin's decentralized structure has reduced reliance on banks and enabled direct and secure financial transactions. Companies like Microsoft, Tesla, Visa, and AXA have begun accepting Bitcoin as a payment method, enhancing its daily use and boosting its popularity. Despite its popularity, the volatility of cryptocurrencies like Bitcoin complicates price predictions. Predicting Bitcoin's price direction is crucial for investors who often rely on fundamental analysis, news tracking, or technical analysis.

This thesis explores the multifaceted impact of Bitcoin price movements in the cryptocurrency market. It leverages statistical models known for their reliability in financial markets and prefers machine learning models for their data compatibility. The study aims to develop a robust model for the dynamics of Bitcoin and other cryptocurrencies, serving as a foundation for future research. It focuses on identifying trend changes and conducting binary predictions on whether prices will rise or fall. The analysis uses Bitcoin data collected daily from the 'Binance' API between August 17, 2017, and January 3, 2024, covering metrics like closing prices, daily highs, and trading volumes. Additionally, economic and financial indicators from other APIs have been integrated.

The study employs time series-specific five-fold cross-validation for training and testing data. It utilizes Generalized Additive Models (including LOESS Logistic Regression and Smoothened Logistic Regression) and machine learning algorithms like Decision Trees, Support Vector Machines, and Gradient Boosting. The models were evaluated based on precision, sensitivity, specificity, F1 score, and accuracy rate, with the Smoothened Logistic Regression model performing best. This model has provided predictions for recent Bitcoin price trends, offering valuable insights for developing financial technologies and investment strategies.

Analyzing the price dynamics in the cryptocurrency market is crucial for enhancing the effectiveness of prediction models and deepening market understanding. The dynamic nature of cryptocurrency markets makes model predictions useful for financial analysis and investment strategies. However, these predictions should be meticulously evaluated in the decision-making process, considering market risks and the limitations of the models.