alors peut être ????

predictions.py

@@ -19,7 +19,7 @@
     from sklearn.ensemble import RandomForestClassifier
     from sklearn.neighbors import KNeighborsClassifier
     from sklearn.tree import DecisionTreeClassifier
-    from sklearn.metrics import accuracy_score, f1_score
+    from sklearn.metrics import classification_report, f1_score
     # from sklearn.model_selection import StratifiedKFold
 except ImportError:
     print("You need to install scikit-learn")
@@ -102,11 +102,14 @@
     model = models[name]
     param_dist = param_distributions[name]
 
+    def scoringfunc(estimator, X, y):
+        return f1_score(y, estimator.predict(X))
+
     randomized_search = RandomizedSearchCV(
         estimator=model,
         param_distributions=param_dist,
         n_iter=NUM_TRIALS,
-        scoring='f1',
+        scoring=scoringfunc,
         cv=3,
         random_state=42,
         n_jobs=-1,
@@ -119,18 +122,19 @@
     model.set_params(**best)
     model.fit(X_train, y_train.values.ravel())
 
-    train_accuracy = accuracy_score(y_train, model.predict(X_train))
+    if DEBUG:
+        print(classification_report(y_test, model.predict(X_test)))
+
     f1_test = f1_score(y_test, model.predict(X_test))
 
     signature = infer_signature(X_train, model.predict(X_train))
 
     with mlflow.start_run(run_name=name) as run:
         mlflow.log_params(best)
         mlflow.sklearn.log_model(model, "model", signature=signature)
-        model_uri = mlflow.get_artifact_uri("model")
-
+
         mlflow.evaluate(
-            model=model_uri,
+            model="runs:/" + run.info.run_id + "/model",
             data=evalData,
             targets='label',
             model_type='classifier',

test.py

@@ -4,9 +4,11 @@
 import random
 from sklearn.model_selection import train_test_split
 from sklearn.preprocessing import StandardScaler
+from sklearn.metrics import classification_report, f1_score
 
 mlflow.set_tracking_uri("https://mlflow.docsystem.xyz")
-model = mlflow.pyfunc.load_model(model_uri="models:/attackdetection/latest")
+# load each models in the RandomSearch experiment
+model = mlflow.sklearn.load_model("runs:/12498379aa574e01b6da4cff313479dd/model")
 
 LABELS_NUM = ["BENIGN", "SUS"]
 
@@ -27,6 +29,21 @@
 X_train = pd.DataFrame(X_train, columns=features.columns)
 X_test = pd.DataFrame(X_test, columns=features.columns)
 
+X_train = X_train.reset_index(drop=True)
+y_train = y_train.reset_index(drop=True)
+X_test = X_test.reset_index(drop=True)
+y_test = y_test.reset_index(drop=True)
+
+
+# calculate the model's metrics
+def calculate_metrics():
+    y_pred = model.predict(X_test)
+    print(classification_report(y_test, y_pred))
+
+    print(f1_score(y_test, y_pred))
+
+calculate_metrics()
+
 good_predictions, bad_predictions = 0, 0
 
 def predict_row(row: int):
@@ -41,7 +58,7 @@ def predict_row(row: int):
 
     return LABELS_NUM[y_test.iloc[row].values[0]] == prediction
 
-while True:
+while False:
     for i in tqdm(range(10000)):
         j = random.randint(0, len(X_test) - 1)
         is_good_prediction = predict_row(j)