Human-Activity-Recognition

It can detect the human activity only from smartphone sensor data.

Dataset

Dataset has been obtained from kaggle.

Classifier

Classifier used is SVM with linear kernel, which showed the best accuracy of 96.33%.

Feature Selection

Variance Threshold is used for feature selection which reduced the accuracy of results, explaining that each of the feature is important for the detection.

Feature Engineering

PCA and KernelPCA are used for feature engineering which also showed no improvements in the results.