Pricing Behavior and Liquidity in the Cryptocurrency Market An Empirical Analysis with Reference to Bitcoin and Altcoins

Abstract

The thesis investigates the pricing dynamics and liquidity in the cryptocurrency market, focusing on Bitcoin and select Altcoins. It examines exchange price trends, forecasts future exchange prices, identifies the underlying factors of price formation, highlights discrepancies across various cryptocurrency exchanges, and compares liquidity of these exchanges with traditional stock exchanges. Existing cryptocurrency market literature reveals several challenges arising from significant volatility and nonlinear price dynamics. Traditional econometric and statistical methods often encounter limitations during extreme market cycles prevalent in the cryptocurrency market, owing to their inherent linear assumptions in an essentially complex market structure. While past literature suggests that machine learning and deep learning models enhance forecasting accuracy, they suffer from a lack of transparency and explainability. In the first objective of this thesis, we address these challenges by analyzing the statistical properties of cryptocurrency price trends, assessing their temporal evolution, and examining the nature of data across various market periods. Before proceeding to data preprocessing, we utilize a high-dimensional multivariate dataset to broaden the range of factors potentially influencing the prices. Subsequently, given the intricacies observed within the data, we introduce a distinctive data preprocessing mechanism. We use a combination of signal-processing techniques to enhance data quality for effective modeling. A novel three-step feature selection method is proposed within this stage, ensuring efficient dimension reduction. The primary aim of this objective is to develop a highly accurate forecasting model that is robust to different market phases and varied short-term forecasting horizons. Accordingly, this research introduces a flexible and explainable financial forecasting architecture that ensures that the resulting predictions are interpretable and tailored for volatile asset market phases.