The Final Results!!!

After training, validating and testing the models available in the Classification Learner app, the LSTM model, and the NPR tool model, it was seen that the Cubic Support Vector Machine provided the best test accuracy of 99.8% for 100% load.  

The table below shows the accuracies obtained for each of the models for 12.5% of the load.

Model	Test Accuracy	Train Accuracy	ROC (AUC)
LSTM	93.04%	92.69%	-
SDNN	96.9%	96.6%	-
Quadratic SVM	96.7%	98.5%	Class 1 – 0.9995
			Class 2 – 0.9999
			Class 3 – 0.9993
			Class 4 – 0.9994
			Class 5 – 0.9994
Cubic SVM	96.6%	98.7%	Class 1 – 0.9993
			Class 2 – 0.9998
			Class 3 – 0.9986
			Class 4 – 0.9992
			Class 5 – 0.9995
Medium Neural Network	96.4%	98.4%	Class 1 – 0.9985
			Class 2 – 0.9995
			Class 3 – 0.9980
			Class 4 – 0.9988
			Class 5 – 0.9990
Bi-layered Neural Network	96.4%	97.8%	Class 1 – 0.9983
			Class 2 – 0.9996
			Class 3 – 0.9973
			Class 4 – 0.9982
			Class 5 – 0.9982
Tri-layered Neural Network	96.0%	97.7%	Class 1 – 0.9981
			Class 2 – 0.9996
			Class 3 – 0.9975
			Class 4 – 0.9976
			Class 5 – 0.9979

 The table below shows the accuracies obtained for each of the models for 100% of the load.

Model	Test Accuracy	Train Accuracy	ROC (AUC)
LSTM	98.18%	98.72%	-
SDNN	98.6%	98.5%	-
Cubic SVM	99.8%	99.6%	Class 1 – 0.9994
			Class 2 – 1.0
			Class 3 – 0.9998
			Class 4 – 0.9998
			Class 5 – 0.9997
Wide Neural Network	99.7%	99.5%	Class 1 – 0.9999
			Class 2 – 0.9997
			Class 3 – 0.9987
			Class 4 – 0.9998
			Class 5 – 0.9998
Quadratic SVM	99.6%	99.3%	Class 1 – 0.9998
			Class 2 – 1.0
			Class 3 – 0.9999
			Class 4 – 0.9996
			Class 5 – 0.9998
Bi-layered Neural Network	99.4%	99.3%	Class 1 – 0.9998
			Class 2 – 0.9997
			Class 3 – 0.9995
			Class 4 – 0.9998
			Class 5 – 0.9988
Medium Neural Network	99.4%	99.4%	Class 1 – 0.9999
			Class 2 – 0.9999
			Class 3 – 0.9997
			Class 4 – 0.9997
			Class 5 – 0.9997