Weather-Integrated Recommendations for Food Choices with SHAP Explanation

In the context of today's rapidly evolving technological landscape and the urbanization of Tier 1 and Tier 2 cities, cloud kitchens have emerged as pivotal players, offering quick and convenient meals without the need for dine-in facilities. The demand prediction for the food has become quite tough and could be considered as one of the significant challenges. Cloud kitchens face challenges in managing stock due to the absence of in-house diners, potentially resulting in wasted food. Moreover, adverse weather conditions often hinder the seamless delivery of online. To address these challenges, this project employs machine learning techniques to predict weather patterns and correlate them with food preferences according to that weather. The key algorithms found to deal with this problem were Random Forest, kNN, Decision tree and Naïve Bayes and Support Vector Machine. This serves the first half of the statement and then for recommendation a suitable system was developed. The pivotal role of the system is to reduce the food wastages for cloud kitchens by fostering adaptive strategies for efficient decision making. The key inputs from the market were found to be that 30-40% of the predicted sales in food industry doesn't meet due to various reasons. The ultimate goal is to empower cloud kitchen operations, ensuring adaptive and informed decision-making in response to changing weather dynamics. The union of both of these are explained using Interpretability tool SHAP for a convenient study.