Throughout this project, we set out to understand what really drives whether a restaurant offers delivery, using a combination of real-world restaurant data and weather information. By analyzing thousands of examples, we discovered that the decision to offer delivery isn’t random—it’s shaped by a mix of business choices, location, and even the weather outside.

One of the most important lessons was how much the weather can shape delivery services. Rain and strong winds don’t just slow down drivers—they can actually change whether restaurants offer delivery at all. We saw that when the forecast predicts storms, restaurants and delivery platforms can plan ahead: by adding more drivers or limiting delivery zones, they can keep customers happy and operations running smoothly. In fact, just a small increase in the number of available drivers on rainy days can make a noticeable difference in delivery speed and reliability.




When it came to predicting which restaurants would offer delivery, we put several different machine learning models to the test. Simpler models like Multinomial Naive Bayes and Logistic Regression performed surprisingly well, reaching accuracy levels above 92%. More advanced approaches, such as Decision Trees and AdaBoost, pushed the accuracy even higher—up to nearly 94%. What stood out was that combining models (as AdaBoost does) helped balance the trade-off between catching all delivery restaurants and avoiding false alarms.

A crucial step in our analysis was making sure our models learned fairly from the data. Originally, the data was heavily imbalanced, with far more restaurants not offering delivery than those that did. By using a technique called SMOTE, we were able to balance the training data, ensuring our models didn’t just “play it safe” by always guessing the majority class. This balancing act was key to boosting the performance of every model we tried, especially Support Vector Machines.

Ultimately, this project shows that data-driven insights can make a real difference for both businesses and customers. By understanding the factors that influence delivery—whether it’s the weather, the type of cuisine, or the city—a restaurant or delivery app can make smarter decisions, improve service, and keep customers satisfied even when the skies turn gray. Our journey proves that with the right data and tools, we can bring clarity to complex, everyday questions and help everyone enjoy a better meal, delivered right to their door.
'