Motion Activity Classification

This repository contains a Jupyter notebook for classifying motion capture activities (Boning and Slicing) using machine learning techniques.

Features

• Loads and processes motion capture CSV data
• Extracts and engineers features (RMS, roll, pitch)
• Handles missing values and class labeling
• Aggregates features by time blocks
• Trains and evaluates multiple models (SVM, Random Forest, SGD, MLP)
• Performs hyperparameter tuning and feature selection
• Compares model performance and selects the best classifier

Usage

1. Place your Boning.csv and Slicing.csv files in the working directory.
2. Run the notebook step by step to preprocess data, extract features, and train models.
3. Review the summary tables and observations to understand model performance.

Requirements

• Python 3.x
• pandas
• numpy
• scikit-learn
• scipy
• tabulate

Install dependencies with:

pip install pandas numpy scikit-learn scipy tabulate

Results

• The notebook compares SVM (with and without tuning), Random Forest, SGD, and MLP classifiers.
• RandomForestClassifier achieves the highest accuracy and is selected as the final model.

---

Author: Jesmine Ting Zi Ching (2026)