Machine Learning Algorithms for Transportation Mode Prediction: A Comparative Analysis

Abstract

This study investigated the performance of various machine learning algorithms in predicting transportation modes from large datasets. The investigated algorithms include Multilayer Perceptron (MLP), K-Nearest Neighbors (KNN), Decision Tree, Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), and Logistic Regression. We rigorously evaluated each algorithm's performance using a robust set of metrics such as precision, recall, and F1-score. This study comprehensively explains the algorithm's capabilities, strengths, and potential weaknesses across seven transportation categories: 'walk', 'bike', 'bus', 'car', 'taxi', 'train', and 'subway'. The Decision Tree (DT) model consistently outperformed the others, demonstrating superior accuracy and a better balance of precision and recall across all modes of transportation. Specifically, it achieved precision, recall, and F1 scores of around 83\% to 94\% across all categories. These findings underline the suitability of the DT model for this classification task and its potential for further applications in transportation mode prediction based on large datasets. However, other algorithms like LSTM and RNN also showed promising results in certain categories, suggesting the value of continued exploration of different models depending on specific use cases.