Hey there! I recently worked on a project to predict Uber ride fares as part of a company interview last year. Instead of relying on a single model, I decided to build a stacking ensemble with several of my top-performing models to improve the results. And the outcome? My final meta-model achieved a mean absolute error (MAE) of 1.2306 on the test set.
I was curious to see how combining multiple models could enhance the prediction accuracy. By stacking these diverse models, I could leverage their strengths and weaknesses to create a more robust predictor. It was a great learning experience, and I’m excited to share my approach with you.
If you’re interested in exploring the project further, I’ve made the full notebook available on GitHub. I’d love to hear your thoughts on alternative approaches you might have taken. Have you worked on a similar project or have any experience with stacking ensemble models?
Let’s discuss!