Transport Mode Detection from Sensor Data

This project focuses on detecting the user's mode of transport (Car, Still) using smartphone sensor data.
It involves two key stages:

• Pre-processing and feature extraction from raw sensor data.
• Training Machine Learning models for transport mode classification.

The project includes Python scripts for data preprocessing and a Jupyter Notebook for model training and evaluation.

---

Dependencies

Make sure you have the following installed:

• Python 3.7+
• NumPy
• Pandas
• SciPy
• Scikit-learn
• Matplotlib (for evaluation plots)

Install required libraries using:

pip install numpy pandas scipy scikit-learn matplotlib

Documentation

Code

In this section we show the functionalities developed in our work and the relative parameters used.

TMDataset.py

Module name	Parameter	Description
load_data(data -dir)	data_dir (str): path to directory containing raw .csv files	Loads all ride data into a single pandas DataFrame.
clean_data(df)	df (DataFrame): raw ride data	Removes missing values and outliers using Z-score method.
compute_orientation(df)	df (DataFrame): cleaned ride data	Calculates roll, pitch, game rotation vector, and orientation using a simplified complementary filter.
calculate_statistics(df)	df (DataFrame): ride data with orientation	Generates statistical features per ride (mean, min, max, std) for each sensor.
main()		Main execution flow: load data, clean, engineer features, and save to a standardized CSV.

Output

The preprocessing script generates:

• standardized_ride_data.csv:
  A clean and feature-engineered dataset ready for model training.

Model Training Code (TransportModeDetection.ipynb)

The Jupyter Notebook covers the complete machine learning pipeline:

• Loading the standardized dataset (standardized_ride_data.csv)
• Splitting the dataset into training and testing sets
• Model Training using:
  • Decision Tree Classifier
  • Random Forest Classifier
  • Support Vector Machine (SVM)
• Model Evaluation using:
  • Accuracy Score
  • Confusion Matrix
  • Classification Report (Precision, Recall, F1-Score)

The notebook demonstrates end-to-end training and testing for transport mode classification based on smartphone sensor data.

How to Run

Step 1: Preprocess Raw Data

Place your ride data .csv files in a folder (e.g., ride_data/).

Then execute the preprocessing script:

python preprocessing.py

This will generate the standardized_ride_data.csv file.

---

Step 2: Train and Evaluate Models

Open the Jupyter Notebook:

jupyter notebook TransportModeDetection.ipynb

Follow the steps inside the notebook to:

• Load the dataset
• Train machine learning models
• Evaluate model performance

###Project Structure

.
├── ride_data/                     # Folder containing raw ride data (.csv files)
├── standardized_ride_data.csv      # Output dataset after preprocessing
├── preprocessing.py                # Python script for data preprocessing
├── TransportModeDetection.ipynb    # Jupyter Notebook for model training and evaluation
└── README.md                       # Project documentation

License

This work is licensed under a MIT License.

Team of collaborators

Lakshman Navaneetha Krishnan and Roshan