the precision and F1 Score of Kettle is keep nan

OdysseasKr / online-nilm

Code for the experiments in "Sliding Window Approach for Online Energy Disaggregation Using Artificial Neural Networks"

60 stars 16 forks source link

the precision and F1 Score of Kettle is keep nan #8

Open Hessen525 opened 4 years ago

Hessen525 commented 4 years ago

Hi @OdysseasKr The precision and F1 Score of Kettle is keep nan all the time. result as following: ============ Recall: 0.0 ============ Precision: nan ============ Accuracy: 0.9962920046349942 ============ F1 Score: nan ============ Relative error in total energy: 0.2103923186958998 ============ Mean absolute error(in Watts): 24.667154793280716

OdysseasKr commented 4 years ago

Hi! This usually means that the model does not predict any activations. In other words, the predicted signal is always below the power on threshold.