Github Link for this project

In this project, extreme gradient boosting (XGBoost) classifier was trained after feature extraction with principal component analysis (PCA) in order to classify human activities. UCI Human Activity Recognition Using Smartphones Data Set was employed for training, and here are the accuracies after 10-times-10-fold cross validation:

Best Accuracy: 99.66%
Mean Accuracy: 95.39%
Var of Accuracy: 9.81%

Keywords about the project:

• Human Activitiy Recognition
• Machine Learning
  • Feature Extraction
  • Principal Component Analysis (PCA)
  • Extreme Gradient Boosting (XGBoost)
  • Cross-validation
  • Confusion Matrix
  • Correlation Heatmaps
  • Fine-tuning
• Python
  • sklearn
  • matplotlib
  • numpy
  • seaborn

More Information

To learn more about the project, you can read the report Human Activitiy Recognition Using XGBoost and PCA.