• Harshal Kuldharan
• Prof. Swati Kadu
• Siddhesh Kulthe
• Atharva Kakade
• Yash Gavhane

Reinforcement Learning (RL) has proven to be a powerful paradigm for sequential decision-making in high-dimensional, uncertain environments, and a natural choice for algorithmic stock trading. In this paper, we synthesize theoretical contributions and empirical methods of recent RL-based stock trading, providing a critical and unified view of current approaches. We discuss cascaded models such as LSTM-RL hybrids, policy-gradient ensembles, and AlphaStock-style deep hierarchical reinforcement learners, as well as root Q-learning models specifically designed for asset allocation. Our combined review centers on important issues such as environment representation, state abstraction, reward design, risk-awareness, and trade execution policy. We also discuss notable challenges such as non-stationarity, exploration-exploitation trade-offs, and market impact modeling. We follow, through this unification, the development of RL trading systems from simple tabular Q learning to sophisticated actor-critic ensembles and provide insights on their weaknesses and strengths and potential future directions, both from an academic and financial industry perspective.

Twin Delayed Deep Deterministic Policy Gradient (TD3), Adam Optimizer, Stock Price Prediction, Deep Reinforcement Learning (DRL), Investment Optimization, Financial Markets

"Stock Prediction using TD3 Algorithm with Adam Optimizer", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 5, page no.k428-k432, May-2025, Available :http://www.jetir.org/papers/JETIR2505B31.pdf

"Stock Prediction using TD3 Algorithm with Adam Optimizer", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 5, page no. ppk428-k432, May-2025, Available at : http://www.jetir.org/papers/JETIR2505B31.pdf