add ap, acc, precision, recall, f1

recbole/evaluator/metrics.py

@@ -27,7 +27,7 @@
 import numpy as np
 from collections import Counter
 from sklearn.metrics import auc as sk_auc
-from sklearn.metrics import mean_absolute_error, mean_squared_error
+from sklearn.metrics import mean_absolute_error, mean_squared_error, average_precision_score, accuracy_score, f1_score, precision_score, recall_score
 
 from recbole.evaluator.utils import _binary_clf_curve
 from recbole.evaluator.base_metric import AbstractMetric, TopkMetric, LossMetric
@@ -380,6 +380,61 @@ def metric_info(self, preds, trues):
         return result
 
 
+class AP(LossMetric):
+    def __init__(self, config):
+        super().__init__(config)
+
+    def calculate_metric(self, dataobject):
+        return self.output_metric("ap", dataobject)
+
+    def metric_info(self, preds, trues):
+        return average_precision_score(trues, preds)
+
+
+class ACC(LossMetric):
+    def __init__(self, config):
+        super().__init__(config)
+
+    def calculate_metric(self, dataobject):
+        return self.output_metric("acc", dataobject)
+
+    def metric_info(self, preds, trues):
+        return accuracy_score(trues, preds > 0.5)
+
+
+class Preci(LossMetric):
+    def __init__(self, config):
+        super().__init__(config)
+
+    def calculate_metric(self, dataobject):
+        return self.output_metric("precision", dataobject)
+
+    def metric_info(self, preds, trues):
+        return precision_score(trues, preds > 0.5)
+
+
+class Recal(LossMetric):
+    def __init__(self, config):
+        super().__init__(config)
+
+    def calculate_metric(self, dataobject):
+        return self.output_metric("recall", dataobject)
+
+    def metric_info(self, preds, trues):
+        return recall_score(trues, preds > 0.5, zero_division=0)
+
+
+class F1(LossMetric):
+    def __init__(self, config):
+        super().__init__(config)
+
+    def calculate_metric(self, dataobject):
+        return self.output_metric("f1_score", dataobject)
+
+    def metric_info(self, preds, trues):
+        return f1_score(trues, preds > 0.5, zero_division=0)
+
+
 # Loss-based Metrics