DEVELOPMENT OF A CRYPTOCURRENCY TREASURY MANAGEMENT SYSTEM

This study aims to develop an innovative solution, supported by next-generation technologies, to facilitate operational integration of crypto assets in banking sector. To this end, a system has been developed that provides real-time risk management and portfolio optimization by combining big data analysis, Natural Language Processing (NLP), and deep learning. The system focuses on price prediction, sentiment analysis, risk measurement, and dynamic portfolio management. In addition, social media and news feeds have been digitized to create an extra layer that provides direct input to Artificial Intelligence (AI) engine. Furthermore, a Deep Reinforcement Learning (DRL) agent has been designed to maximize risk-adjusted return instead of absolute return, to adapt to stochastic nature of market. Deterministic risk models have been integrated with proactive safety mechanisms. This allows for focusing on financial risk, operational risk, and safety dimensions. A hybrid method combining Long Short-Term Memory (LSTM) and Transformer models has been proposed to simultaneously capture linear and nonlinear market dynamics. Auto Regresive Integrated Moving Average (ARIMA) and LSTM-based prediction models have also been developed to analyze prediction success of proposed method in this study. Examination of results revealed that proposed hybrid model exhibited high performance.