This repository contains a set of python modules, classes and notebooks that allow the evaluation of the travel mode detection algorithm proposed in US-TransportationMode using k-fold cross-validation.
It also allows the evaluation of the improvements in detection performance and cost obtained with the use of the Automated Machine Learning (AutoML) framework provided by AutoSklearn and Principal Component Analysis (PCA) class provided by Scikit-Learn
The evaluation experiments were conducted using smartphone collected data from the TMD-Dataset.
The python modules and classes contained in this repository are extended and modified versions of the ones made available in the US-TransportationMode repository. US-Transportation mode was licensed under MIT License and licensing information can be verified in the LICENSE.md file.
The evaluation experiments performed using the jupyter notebooks available in this repository have been submitted for publication in a peer-revied journal. Due to the maximum number of pages that can be included in a journal paper, supplementary tables containing details about the experiments have been made made available in this repository:
-
experiments_config.pdf: This file details the configurations for AutoSklearn and traditional machine learning algorithms used in each evaluation scenario.
-
ensembles_config.pdf: This file details the ensemble configurations generated with AutoSklearn during each evaluation scenario.
-
experiments_results.pdf: This file details the results obtained in each evaluation scenario by the ensembles generated with AutoSklearn and traditional machine learning algorithms.
To install python modules and notebooks dependencies run:
pip install -r requirements.txt
To install AutoSklearn and its dependencies run:
curl https://raw.githubusercontent.com/automl/auto-sklearn/master/requirements.txt | xargs -n 1 -L 1 pip install
pip install auto-sklearn
- Data Visualization.ipynb was used to generate histograms and line plots
- TravelModeDetection-OriginalDataset.ipynb was used to evaluate the scenarios in which ONLY the original features of the dataset were used, with or without PCA.
- TravelModeDetection-ModifiedDataset.ipynb was used to evaluate the scenarios in which the Skewness and Kurtosis features were used in addition to the ones contained in the original dataset.