Ashok Jana

Financial markets don’t sit still. Volatility spikes, trends reverse, and what worked last month stops working today. This project builds an adaptive trading agent using deep reinforcement learning to handle these shifts. Instead of fixed rules or static predictions, the agent learns trading policies through simulated market interaction. I’ll use deep neural networks to approximate action-value functions for sequential decisions. Raw historical data gets processed into state representations capturing price movements, volume, volatility signals, and macro indicators. Several baseline strategies get compared against different RL configurations to test what actually improves risk adjusted returns. The core question: can agents learn to recognize and adapt to different market regimes? This project tests whether regime-aware representations help models generalize across bull markets, bear markets, and choppy sideways action.

The quick development of digital media and social networking sites have greatly enhanced misinformation and fake news spread. The ease with which news and information is accessed online with millions of users dependent on the online platforms has made it more challenging to determine the credibility of the information published. His or her fake news has the potential to inform the opinions of the people, cause confusion, and affect the political, economic, and social systems. The given project is dedicated to the creation of an automated system of fake news detection based on Natural Language Processing and the Bidirectional Encoder Representations from Transformers (BERT) deep learning model. BERT is a language model that has the ability to comprehend the existence of contextual relationships between words within a sentence, and this is a transformer-based language model. The model can distinguish between fake and real news articles better by analyzing the language patterns and contextual characteristics of a specific news article. The research implies gathering publicly accessible news materials, textual data preprocessing, and training machine learning models to classify them. The suggested system will ensure less misinformation through the automatic detection of misleading news articles. The conclusions of this project could be utilized to make digital information systems more reliable and provide social media with more efficient methods of detecting fake news.