Food Delivery Time Prediction in Indian Cities Using Machine Learning Models Ananya Garg ∗ ananya22068@iiitd.ac.in Mohmmad Ayaan ∗ ayaan22302@iiitd.ac.in Swara Parekh ∗ swara2022524@iiitd.ac.in Vikranth Udandarao ∗ vikranth22570@iiitd.ac.in Department of Computer Science Indraprastha Institute of Information Technology, Delhi ∗ All authors have contributed equally to this work. The code implementation is available at: https : / / github . com / Vikranth3140/Food-Delivery-Time-Prediction. Abstract Accurate prediction of food delivery times significantly impacts customer satisfaction, operational efficiency, and profitability in food delivery services. However, existing studies primarily utilize static historical data and often overlook dynamic, real-time contextual factors crucial for precise prediction, particularly in densely populated In- dian cities. This research addresses these gaps by inte- grating real-time contextual variables such as traffic den- sity, weather conditions, local events, and geospatial data (restaurant and delivery location coordinates) into predic- tive models. We systematically compare various machine learning algorithms, including Linear Regression, Decision Trees, Bagging, Random Forest, XGBoost, and LightGBM, on a comprehensive food delivery dataset specific to Indian urban contexts. Rigorous data preprocessing and feature selection significantly enhanced model performance. Ex- perimental results demonstrate that the LightGBM model achieves superior predictive accuracy, with an R² score of 0.76 and Mean Squared Error (MSE) of 20.59, outperform- ing traditional baseline approaches. Our study thus pro- vides actionable insights for improving logistics strategies in complex urban environments. The complete methodol- ogy and code are publicly available for reproducibility and further research. 1. Introduction The rapid growth of online food delivery services has significantly transformed urban consumption patterns, par- ticularly in Indian cities where platforms like Zomato and Swiggy dominate the market. Providing accurate and re- liable estimates of delivery times is essential not only for enhancing customer satisfaction but also for optimizing operational efficiency and reducing overall delivery costs. However, accurately predicting food delivery times remains challenging due to various uncontrollable and dynamic fac- tors such as traffic congestion, variable weather conditions, and sudden demand fluctuations caused by local festivals or events. Existing research in the domain predominantly relies on static historical data, such as historical average delivery du- rations and past order volumes. These traditional methods often neglect dynamic, context-specific factors like real- time traffic conditions, weather variability, and geographic complexities, which are particularly relevant in the con- text of Indian urban environments. The oversight of these critical variables leads to inaccurate predictions and subse- quently undermines operational performance. In this paper, we explicitly address this research gap by proposing and evaluating a novel machine learning-based predictive framework that leverages real-time contextual and geospatial information. Our approach integrates criti- cal features such as real-time traffic density, weather con- ditions, and geographic distance between restaurants and customer locations, combined with comprehensive demo- graphic and logistical information about delivery personnel and order specifics. To achieve our objectives, we systematically evaluate and compare a range of predictive modeling techniques, in- cluding traditional methods like Linear Regression and ad- vanced ensemble models such as Random Forest, XGBoost, and LightGBM. Through rigorous preprocessing and care- ful feature selection, we demonstrate that the integration of