Akeed Restaurant Recommendation Challenge

An image of the Akeed Food Delivery Restaurant/comapny based in Oman

This is the 30th place solution to the Akeed Restaurant Recommendation Challenge on ZINDI.

The objective of this competition is to build a recommendation engine to predict what restaurants customers are most likely to order from given the customer location, restaurant information, and the customer order history.

Summary of Approach

A single model with optimized parameters gave a public leaderboard score of 0.232

Model Architecture

. LightGBM

. Optimized Parameters

. Thresholding

Data Enginnering & Transformation

. Vectorization

. Time - seconds

. Numerical value fit transformation

. Mean of Longitude and Latitude

. Using Numerical Data as Categorical Data

Common Configuration

. LightGBM Feature importance

. Optimizinlg learning rate

. Using LightGBM categorical feature

. Data Reduction