In this project, I developed a model for classifying exercise types using fitness tracker data collected over a week. The goal was to create an accurate and reliable system for categorizing different types of exercises based on the collected data.
-
Data Pre-processing:
- Conducted comprehensive data pre-processing, including cleaning, merging, and resampling techniques. Ensured the data was ready for further analysis and model training.
-
Outlier Detection:
- Implemented various outlier detection methods, such as Interquartile Range (IQR), Local Outlier Factor, and Chauvenet’s criterion. This step aimed to identify and handle anomalies in the dataset.
-
Feature Engineering:
- Utilized a variety of feature engineering techniques to enhance model performance. This included applying Butterworth’s low-pass filter, Principal Component Analysis (PCA), and Fourier Transformation to extract meaningful features from the data.
-
Feature Selection:
- Employed "Forward Selection" as a feature selection method to identify the most relevant features for model training. This step helped improve model efficiency and reduce dimensionality.
-
Model Training:
- Trained several machine learning models, including Random Forest, Artificial Neural Network (ANN), Naïve Bayes, Decision Tree, and K-Nearest Neighbors (KNN).
-
Model Evaluation:
- Identified Random Forest as the best-performing model with an impressive 98% accuracy. The model demonstrated its effectiveness in predicting exercise types, even on previously unseen data.
This project serves as a foundation for further exploration and improvement. Potential future enhancements may include refining the model architecture, incorporating additional features, and exploring real-time predictions in a fitness tracking application.
I would like to express gratitude for the support and guidance received throughout this project. Collaborative efforts and insightful feedback contributed to the successful development and validation of the exercise classification model.
Feel free to adjust the content to fit the specific details of your project.