Added F3 score check for performance evaluation

model.py

@@ -3,6 +3,7 @@
 import argparse
 import csv
 from sklearn.linear_model import SGDClassifier as SGD
+from sklearn.metrics import fbeta_score
 import numpy as np
 from utils import *
 
@@ -18,16 +19,33 @@ def main():
         next(r) # skip headers
 
         model = SGD(loss="hinge", penalty="l2", max_iter=100)
+        
+        # Model training
+        print("Starting Model Training...")
         train_data = prepare_train_data('data/training.csv') # extract features from input data
-        X_train = train_data.loc[:,'sex':'D'] # input features
-        Y_train = train_data.loc[:,'aki'] # target labels
+        train_split = len(train_data)*0.8 # split training data into training (80%) and validation (20%) sets
+        X_train = train_data.loc[:train_split,'sex':'D'] # input features
+        Y_train = train_data.loc[:train_split,'aki'] # target labels
         model.fit(X_train.values, Y_train.values) # train SGD model
+        print("Model Training Complete.")
 
+        X_test = train_data.loc[train_split:,'sex':'D']
+        Y_test = train_data.loc[train_split:,'aki']
+        test_pred = model.predict(X_test.values)
+
+        f3_score = fbeta_score(Y_test, test_pred, beta=3)
+        if f3_score < 0.7:
+            raise Exception(f"Model performance ({f3_score}) is below the required threshold of 0.7.")
+        else:
+            print(f"Model performance meets F3 score requirements ({f3_score})")
+
+        print("Starting Model Prediction...")
         for row in r:
             patient = prepare_test_data(row) # format input data for model prediction
             y_pred = model.predict(np.array(patient).reshape(1, -1))
             aki_pred = to_label(y_pred[0], 'y', 'n') # convert binary prediction to 'y'/'n' labels
             w.writerow((aki_pred,))
+        print("Model Prediction Complete.")
 
     except Exception as e:
         print(f"An error occurred while running the prediction: {e}")