README.md

Smart Chair

ML classifier and analysis

We used different ML algorithms to train a classifier model to classify a test subject's posture. Here is how different models performed.

Tree

Max number of splits: 100 Split criterion: Gini's diversity index Surrogate decision splits: off

Confusion matrix:

Results:

Support Vector Machine (SVM)

• Linear SVM

Confusion matrix:

Results:

• Gaussian SVM

Confusion matrix:

Results:

K Nearest Neighbours (KNN)

Number of nearest neighbours: 3 Distance metric: Euclidean

Confusion matrix:

Results:

Neural Networks

• network 1:

number of layers: 3 first layer size: 100 second layer size: 25 third layer size: 10 activation: ReLU

Confusion matrix:

Results:

• network 2:

number of layers: 3 first layer size: 50 second layer size: 25 third layer size: 10 activation: ReLU

Confusion matrix:

Results:

• network 3:

number of layers: 3 first layer size: 20 second layer size: 15 third layer size: 10 activation: ReLU

Confusion matrix:

Results:

Note that all the confusion matrices are from validation.

For our use case it's best to have the model with least training time.