• Built a model using the keras library and dpylr package in R to predict the type of activity a person is doing.
• Used the ReLU activation function and trained the network for 100 epochs using batch gradient descent.
• Fitted the trained model to the test dataset and computed the balanced accuracy, per-class sensitivity, and specificity.
• Performed mini batch gradient descent and retrained the model.
• Added a new hidden layer to the network and documented the changes on the performance of the model.