This study examined the price volatility of Indian agricultural commodities caused by natural and policy-driven shocks. Using event study methodology with panel regression, volatility analysis, and anomaly detection, the project analyzed daily data (2014–2025) for key crops like rice, wheat, pulses, and vegetables. Results quantified abnormal returns in response to specific events, isolating direct price impacts. The findings equip policymakers and market participants with evidence to design timely interventions, stabilize markets, and mitigate risks to farmers and consumers. By addressing volatility, the study contributes to improved food security and economic resilience.

This project focused on predicting USD/INR exchange rates to aid policymakers, investors, and corporations. A multi-domain dataset spanning three decades was created, combining economic indicators with sentiment data extracted from financial news using FinBERT. Classical time-series models (ARIMA, VAR) were compared with advanced ML (XGBoost, Random Forest) and Deep Learning (LSTM). Results showed ensemble ML models outperformed others, especially with sentiment integration, capturing nonlinear market dynamics. Key drivers included crude oil, Nifty, and forex reserves. The hybrid forecasting framework offers central banks early warning systems and supports corporate hedging and risk management strategies.

Addressing inefficiencies in FMCG distribution, this project developed predictive models to reduce cancellations, delays, and SLA violations. Logistic regression predicted high- risk orders likely to be canceled or returned, while LightGBM models forecast delivery times and stage-wise delays. By shifting from reactive to proactive decision-making, distributors gained visibility to allocate resources effectively and minimize working capital risks. The scalable framework enhances operational efficiency, reduces financial risk, and supports profitability in the competitive FMCG sector. The study demonstrates the power of analytics in transforming distribution systems into data-driven, resilient operations.

This project developed a data-driven optimization model for a most popular IPL franchise auction strategy. A Most Valuable Player (MVP) index was designed using player metrics like batting strike rate, bowling economy, and past performance. The index became the objective function in an Integer Linear Programming (ILP) model, which considered constraints such as squad size, budget cap (₹1 billion), overseas player limits, and role requirements. The optimized output suggested the most cost- effective squad composition to maximize team performance. This structured framework provides franchises a scientific approach to player acquisition, balancing budget constraints with competitive edge.

With renewable energy growth creating instability in India’s power grid, this project developed predictive models to forecast short-term frequency deviations. Using the CRISP-DM methodology, multiple models including ARIMA, LSTM, GRU, and XGBoost were tested on 30-minute frequency predictions. GRU and XGBoost achieved the best accuracy in handling time dependencies and variability. The model enables generators to adjust production proactively, avoiding costly penalties under CERC regulations. It provides a decision-support tool to improve grid stability, optimize operations, and leverage market opportunities, strengthening resilience in India’s evolving energy landscape.

This project aimed to optimize a world-renowned café giant’s most popular rewards program by predicting customer responses to promotional offers. Using demographic and transactional data, machine learning models evaluated how factors like recency, frequency, and monetary value influence acceptance of discounts and “Buy One Get One” deals. Analysis revealed variations in response by customer segments, allowing for targeted personalization. The predictive framework improves ROI by reducing ineffective campaigns, lowering operational costs, and enhancing loyalty program efficiency. It ensures that the right offers reach the right customers, boosting engagement, repeat purchases, and long-term profitability for Starbucks.

Using NFHS-5 data, this study analyzed the pandemic’s impact on reproductive health and early childhood education in India. Contraceptive access declined significantly, especially in urban areas, as modern methods became less accessible. Logistic regression showed wealth and healthcare decision-making influenced urban contraceptive use. Regional analysis revealed higher women’s empowerment in southern states. Preschool enrollment for ages 2–4 dropped from 46.6% to 30.4%, disproportionately affecting wealthier, urban households reliant on private schools. Findings emphasize the need for targeted policy to restore family planning services and address growing disparities in early childhood education and gender empowerment.

This project tackled flight delays at European airports, especially London Heathrow, by optimizing runway use and resource allocation. A combinatorial optimization model using Gurobi reduced runway occupancy by 6.5 minutes per hour, enabling 5–6 more flights. A second sequential decision optimization framework predicted delays and adjusted resource allocation, such as ground staff scheduling. Together, the models improved on-time performance, minimized congestion, and boosted passenger satisfaction. The research demonstrates how advanced optimization and predictive analytics can provide a practical roadmap for airports to enhance operational efficiency, capacity utilization, and service reliability.

This project examined how online presence impacts restaurant success beyond food quality. Using Yelp data, it combined sentiment analysis of customer reviews with image feature extraction from restaurant photos. Analysis showed that higher-priced restaurants received stronger positive sentiments, particularly about ambiance, and that visual aesthetics like brightness and color contrast influenced ratings. By integrating text and image analytics, the study predicted restaurant longevity and performance more accurately. Findings provide restaurant owners actionable insights to align customer experience, service quality, and visual branding, improving competitiveness and long- term sustainability.

This project developed a multi-model stock recommendation system by integrating fundamental, technical, and sentiment analysis for post-COVID markets. Fundamental indicators assessed financial health, technical analysis captured price patterns, and sentiment models using NLP (BERT, VADER) gauged investor mood. The ensemble of models generated superior buy/sell signals compared to single-method approaches and consistently outperformed the Nifty 50 benchmark. Portfolio rebalancing strategies maintained diversification and adapted to market shifts. This comprehensive framework offers investors a powerful tool for navigating volatility with better-informed decisions and improved portfolio performance.

A southern Indian restaurant chain faced demand uncertainty due to frequent sell-outs and perishable items. This project used Survival Analysis to estimate “true demand” masked by stockouts and built forecasting models (SARIMAX, Prophet, Ensemble) integrating external variables like weather, festivals, and auspicious days. The improved forecasts reduced wastage and optimized production at the central kitchen. Accurate demand planning improved efficiency, reduced costs, and enhanced profitability across outlets. The solution demonstrates the importance of integrating contextual external factors into predictive models for more reliable demand forecasting in the food industry.

This project proposed a dynamic premium calculation model for usage-based insurance (UBI). Traditional models rely on static factors like age and mileage, but this framework incorporated real-time driving behavior, emphasizing fatigue and drowsiness as major accident risks. By integrating behavioral metrics and drowsiness detection systems, insurers can better assess risk and incentivize safer driving through tailored pricing. The approach improves accuracy in premium determination while promoting road safety. It provides insurers a competitive edge, reducing claims while rewarding safe drivers, and helps transform UBI into a more personalized and effective system.

This project enhanced factor-based investment strategies by combining macroeconomic indicators with investor sentiment. Using NLP to analyze sentiment from RBI monetary policy announcements, the study integrated these insights into multi-factor models on Nifty 100 stocks. Portfolios based on sentiment shifts and actual-versus-expected policy rate changes outperformed benchmarks in backtesting. The strategy demonstrates how investor psychology, when combined with fundamentals, can generate alpha and exploit market inefficiencies. It equips fund managers with a robust framework to improve investment returns and adapt to the growing role of sentiment in modern financial markets.

Targeting social media influencers, especially on LinkedIn, this project developed an AI- powered content creation tool to optimize engagement. The system analyzes follower demographics, behavior, and engagement history, then generates tailored content. By leveraging advanced analytics, it identifies themes most likely to resonate with specific audiences. The tool provides influencers with a structured approach to content personalization, boosting engagement rates, reach, and follower loyalty. It helps transform inconsistent content performance into predictable outcomes, empowering influencers and marketers to refine their strategies and maximize the ROI of their social media campaigns.