init commit

lab4/train_spam_model.py

@@ -0,0 +1,75 @@
+import joblib
+import csv
+
+import numpy as np
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.linear_model import LogisticRegression, Ridge
+from sklearn.metrics import (
+    accuracy_score,
+    classification_report,
+    mean_absolute_error,
+    r2_score,
+)
+from sklearn.model_selection import train_test_split
+
+
+def load_data(path="data.csv"):
+    texts, labels, scores = [], [], []
+    with open(path, newline="", encoding="utf-8") as f:
+        for row in csv.DictReader(f):
+            texts.append(row["text"])
+            labels.append(int(row["label"]))
+            scores.append(float(row["spam_score"]))
+    return texts, np.array(labels), np.array(scores)
+
+
+def main():
+    texts, labels, scores = load_data()
+
+    X_train_txt, X_test_txt, y_train, y_test, s_train, s_test = train_test_split(
+        texts, labels, scores, test_size=0.25, random_state=42, stratify=labels
+    )
+
+    vectorizer = TfidfVectorizer(lowercase=True, ngram_range=(1, 2), max_features=2000)
+    X_train = vectorizer.fit_transform(X_train_txt)
+    X_test = vectorizer.transform(X_test_txt)
+
+    # Classification Model (Spam/Ham)
+    clf = LogisticRegression(max_iter=1000)
+    clf.fit(X_train, y_train)
+    y_pred = clf.predict(X_test)
+
+    print("Classification: Spam vs Ham")
+    print(f"Accuracy: {accuracy_score(y_test, y_pred):.3f}")
+    print(classification_report(y_test, y_pred, target_names=["ham", "spam"]))
+
+    # Regression model
+    reg = Ridge(alpha=1.0)
+    reg.fit(X_train, s_train)
+    s_pred = reg.predict(X_test)
+    s_pred = np.clip(s_pred, 0, 1)
+
+    print("Regression: Spam Score (0-1)")
+    print(f"MAE: {mean_absolute_error(s_test, s_pred):.3f}")
+    print(f"R^2: {r2_score(s_test, s_pred):.3f}")
+
+    # Demo model
+    print("\nTest Samples")
+    samples = [
+        "Congratulations! You've won a free vacation, click here now!",
+        "Hey, can you send me the meeting notes from earlier?",
+    ]
+    X_demo = vectorizer.transform(samples)
+    preds = clf.predict(X_demo)
+    scores_demo = np.clip(reg.predict(X_demo), 0, 1)
+    for text, label, score in zip(samples, preds, scores_demo):
+        verdict = "SPAM" if label == 1 else "HAM"
+        print(f"[{verdict}] score={score:.2f} :: {text}")
+
+
+if __name__ == "__main__":
+    # persist our model
+    joblib.dump(vectorizer, "vectorizer.joblib")
+    joblib.dump(clf, "classifier.joblib")
+    print("\nSaved model artifacts.")
+    main()