Using Machine Learning for Performance Classification and Early Fault Detection in Solar Systems

Table 2

Experimental results on the solar dataset using different ML models.

ML algorithm	Accuracy	F-monitoring	F-inspecting	F-running	Weighted avg. F-score	Weighted avg. precision	Weighted avg. recall
ZeroR	45.65%	0.627	0	0	0.217	0.457	0.545
Random Forest	98.28%	0.981	0.971	0.993	0.983	0.983	0.983
J48	98.85%	0.988	0.978	0.997	0.989	0.989	0.989
LibLinear	82.95%	0.801	0.832	0.863	0.828	0.833	0.830
Naïve Bayes	91.18%	0.904	0.908	0.926	0.912	0.915	0.912
Linear Regression	97.25	0.970	0.939	0.998	0.973	0.973	0.973
CNN	91.42%	0.903	0.920	0.925	0.914	0.915	0.914

Values highlighted in bold indicate the highest value across all models with respect to a certain metric.