Badrikanath Praharaj
Research Internship @Vit Bhopal University
Signup · Get unlimited contacts
WORK HISTORY
Research Internship @Vit Bhopal University
Model FrameworkThe paper introduces an Adaptive ARIMA-HMM forecasting model tailored to the Indian stock market (S&P BSE Sensex). The approach combines ARIMA(1,0,1) for capturing short-term linear dynamics with a 3-state Gaussian Hidden Markov Model (HMM) applied to ARIMA residuals for identifying latent market regimes. These regimes—Bull, Bear, and Sideways—are decoded using the Viterbi algorithm, while smoothed regime probabilities enable dynamic allocation of portfolio weights (100% in Bull, 50% in Sideways, 0% in Bear). This regime-aware strategy allows for more stable and risk-sensitive investment decisions compared to static models.Empirical ResultsExtensive experiments highlight the superiority of the adaptive hybrid over baseline models. While standalone ARIMA achieved a total return of 141.41% with a Sharpe ratio of 0.56, and traditional ARIMA-HMM underperformed with negative returns, the proposed adaptive version achieved an extraordinary total return of 3040.88%, annualized return of 40.44%, and a Sharpe ratio of 4.63, all while reducing maximum drawdown to 19.62%. Sensitivity analysis confirmed that a 3-regime structure strikes the optimal balance between profitability and stability, outperforming 2-state (underfitting) and 5-state (overfitting) variants.SignificanceBy integrating regime probabilities into forecasting and allocation, the Adaptive ARIMA-HMM addresses limitations of linear-only and static hybrid models. The framework not only captures market transitions but also optimizes equity exposure to minimize risk during downturns and capitalize on bullish phases. This makes it a powerful quantitative finance tool for investors and policymakers in emerging markets, where volatility, policy shifts, and structural breaks demand adaptive and data-driven forecasting strategies.
EDUCATION
Vellore Institute of Technology
Bachelor of Technology - BTech
ABOUT BADRIKANATH PRAHARAJ
Ever since I built my first robot in high school, I\'ve been passionate about creating technology that makes a real difference. As a third-year student at VIT Bhopal specializing in AI and robotics, I focus on practical and affordable solutions.Inspired by farmers struggling with weeds, I developed an Automatic Weed Detection System using AI to identify weeds with 95% accuracy, saving time and reducing manual labor. I led a team to create a Weather Monitoring System that costs 90% less; it\'s now used in three villages, helping over people access timely weather updates. Our 6 DOF Robotic Arm project made advanced robotics five times cheaper.In research, I\'ve co-authored papers on predicting stock prices, detecting diseases, and recognizing sign language, all achieving over 99% accuracy. My work on detecting Lumpy Skin Disease in cattle aims to improve animal healthcare.As Co-Lead of the R&D Team at VIT, I mentor younger students and foster collaboration. Skilled in Python, C++, and Java, I use TensorFlow and PyTorch to tackle challenges.Believing technology should serve everyone, I aim to create accessible AI solutions and am excited to connect with others who share this vision. If you\'re interested in collaborating, let\'s get in touch!
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.