Multi-Objective Optimization Using Deep Reinforcement Learning for Food E-Commerce Industry

Abstract

In this study, we try and come up with an optimal solution for a given set of conflicting multiple objectives, it's in human nature to get the maximum out of a situation, which involves optimization of more than just a single objective at a time, thus we propose a model which can optimize multiple conflicting objectives and provide the best solution which has less trade-off between the objectives. Deep Reinforcement Learning (DRL) for Multi-Objective Optimization (MOO), deep reinforcement learning combines deep learning artificial neural networks with a reinforcement learning agent that uses a trial-and-error method to reach the goal. In this study we mainly consider the food delivery system and work with the Zomato dataset and optimize objectives such as cost per person, estimated time of arrival (ETA), and ratings of the restaurant, we try and find out the best-fit restaurants for the user based on his/her objectives at the time, we imbibe deep-reinforcement learning to achieve the task, by doing so we try to provide better user experience and contribute to the food e-commerce industry.