MTRecS-DLT: Multi-Modal Transport Recommender System using Deep Learning and Tree Models

Abstract

Multi-modal transport recommender systems aim to provide different users with different route choices for more than one mode of transportation. Most existing systems focus on unimodal transportation providing shortest distance or travel time. Knowing that the use of machine learning and deep learning techniques are achieving success in many fields, it has also been applied to improve the transport networks by helping individuals to meet their needs and observe their various preferences. In this paper, we develop a model called MTRecS-DLT (Multi-Modal Transport Recommender System using Deep Learning and Tree Models) for recommending the most appropriate transport mode for different users. We have used the weighted average ensembling method of Convolutional Neural Network (CNN) and Gradient-Boosted Decision Trees (GBDT) that shows promising results. We have extracted context and user features from the training data. Then, CNN has been applied to extract latent features. The proposed model utilizes a weighted average ensembling to combine CNN and GBDT.