LSTM Network for Transportation Mode Detection

Abstract

The study of Transportation Mode Detection (TMD) has become a popular research field in recent years. It will be a crucial part of Smart mobility and Smart cities in upcoming years. In our study, using the approach of TMDataset, we have gathered the data from different user’s Smartphones up to 5 different transportation modes. However, as the raw data contains noise, we use Feature Engineering to extract useful features from the raw dataset and convert it into different feature frames to feed into a deep learning model called Long Short-Term Memory (LSTM). We used different sized feature frames to input the LSTM network for efficient transportation mode detection and achieved up to 98% classification accuracy for five transportation modes.