This repository contains code for Human Activity Recognition (HAR) using the ESENSE wearable sensor. We use two different methods to validate our models: Leave-One-Out cross-validation and Random Split.
- Python 3.x
- Scikit-learn
- NumPy
- pandas
We use the Dataset for Leave-One-Out cross-validation and my-data for random-split for Human Activity Recognition.
We use the following two methods for model validation:
In this method, we use all but one of the data instances to train our model, then validate the model on the left out data instance. This process is repeated such that each data instance is used exactly once for validation.
In this method, we randomly split our data into a training set and a test set. The model is trained on the training set, and validated on the test set.
The performance of our model was as follows:
-
[INFO]: Performance metrics when using Leave-One-Out Cross-Validation
-
RANDOM FOREST
-
SUPPORT VECTOR
-
MLP CLASSIFIER
-
LOGISTIC REGRESSION
-
[INFO]: Performance metrics when using Random Split
-
- RANDOM FOREST
-
SUPPORT VECTOR