AI-Driven Portfolio Optimization: Integrating Sentiment Analysis, Reinforcement Learning, and Personalised Advisory

Abstract: Traditional portfolio optimization methods, such as Mean-Variance Optimization (MVO) and the Capital Asset Pricing Model (CAPM), rely on static historical data and linear assumptions about asset returns and correlations, making them ineffective during volatile market conditions, when relationships between assets break down and historical patterns fail to generalise. These classical approaches also cannot incorporate unstructured data sources such as news sentiment, social media discourse, or macroeconomic indicators, leaving valuable market signals untapped. This paper presents FinanceWiz, an integrated AI-powered portfolio optimization platform that combines LSTM-based return prediction, sentiment analysis from news and social media, reinforcement learning for dynamic rebalancing, and a personalised AI investment coach. The system was evaluated on 10 years of multi-asset data across equities, forex, commodities, and cryptocurrencies. The AI advisory module showed that it could respond to different situations. Our method works when the market is very volatile and gives better risk adjusted returns than Mean Variance Optimizer (MVO) baseline. Keywords: Portfolio Optimization, Artificial Intelligence, Machine Learning, Reinforcement Learning, Risk Management, Sentiment Analysis, Investment Advisory. Title: AI-Driven Portfolio Optimization: Integrating Sentiment Analysis, Reinforcement Learning, and Personalised Advisory Author: Sana Perween, Aayush Rathi, Tanush Vora International Journal of Computer Science and Information Technology Research ISSN 2348-1196 (print), ISSN 2348-120X (online) Vol. 14, Issue 2, April 2026 - June 2026 Page No: 23-30 Research Publish Journals Website: www.researchpublish.com Published Date: 17-April-2026 DOI: https://doi.org/10.5281/zenodo.19627138 Paper Download Link (Source) https://www.researchpublish.com/papers/ai-driven-portfolio-optimization-integrating-sentiment-analysis-reinforcement-learning-and-personalised-advisory