scoring function improved

pvpower/estimator.py

@@ -207,7 +207,7 @@ def num_samples_last_train(self) -> int:
     def retrain(self, train_data: TrainData):
         start = time.time()
         samples = train_data.samples
-        samples = [sample for sample in samples if sample.irradiance > 0]  # do not train the estimator for zero irradiance
+        samples = [sample for sample in samples if sample.irradiance > 0]  # special handling zero irradiance records
         feature_vector_list = [self.__vectorizer.vectorize(sample) for sample in samples]
         label_list = [sample.power_watt for sample in samples]
         if len(set(label_list)) > 1:
@@ -223,7 +223,7 @@ def retrain(self, train_data: TrainData):
             logging.debug("estimator can not be trained. Insufficient train data")
 
     def predict(self, sample: WeatherForecast) -> int:
-        if sample.irradiance > 0:   # shortcut. no irradiance means no power
+        if sample.irradiance > 0:   # special handling zero irradiance records. no irradiance means no power
             if self.__num_samples_last_train < 1:
                 logging.warning("estimator has not been trained (insufficient train data available). returning 0")
                 return 0

pvpower/trainingcenter.py

@@ -28,7 +28,7 @@ def score(self) -> float:
                 continue
             scores.append(self.__score(real, predicted))
         scores_without_outliners = self.__without_outliners(scores, 0.1)
-        return mean(scores_without_outliners)
+        return round(mean(scores_without_outliners), 2)
 
     @staticmethod
     def __without_outliners(scores: List[int], percent: float) -> List[int]: