Update app.py

app.py

@@ -100,36 +100,46 @@ import tensorflow as tf
 import pickle
 
 # ---------------- 載入模型 ----------------
-model = tf.keras.models.load_model("AIDetect.h5")  # 你的模型檔案
+model = tf.keras.models.load_model("AIDetect.h5")
 with open("vectorizer.pkl", "rb") as f:
     vectorizer = pickle.load(f)
 with open("scaler.pkl", "rb") as f:
     scaler = pickle.load(f)
 
-# ---------------- 特徵計算（純 Python） ----------------
+# ---------------- 純 Python 特徵計算 ----------------
 def compute_features(text):
     words = text.split()
     word_count = len(words)
-    unique_word_ratio = len(set(words)) / (word_count + 1e-6)
+    unique_words = len(set(words))
+    unique_word_ratio = unique_words / (word_count + 1e-6)
     repeat_rate = 1 - unique_word_ratio
     punctuation_count = sum(1 for c in text if c in ".,!?;:")
     punctuation_ratio = punctuation_count / (len(text) + 1e-6)
-    avg_word_length = sum(len(w) for w in words) / (word_count + 1e-6) if words else 0
-    # 直接返回列表，不用 numpy
+    avg_word_length = sum(len(w) for w in words) / (word_count if word_count else 1)
     return [[word_count, unique_word_ratio, repeat_rate, punctuation_ratio, avg_word_length]]
 
+# ---------------- 純 Python 標準化 ----------------
+def transform_features(feat):
+    # scaler 是舊的 scikit-learn StandardScaler，裡面有 mean_ 和 scale_
+    mean = scaler.mean_
+    scale = scaler.scale_
+    transformed = []
+    for i, val in enumerate(feat[0]):
+        transformed.append((val - mean[i]) / scale[i])
+    return [transformed]
+
 # ---------------- 生成解釋 ----------------
 def explain_prediction(text):
     # 文字向量化
     seq = vectorizer([text])
-    seq = tf.keras.preprocessing.sequence.pad_sequences(seq, maxlen=50, padding='pre')
+    seq = tf.keras.utils.pad_sequences(seq, maxlen=50, padding='pre')
 
     # 統計特徵
     feat = compute_features(text)
-    feat = scaler.transform(feat)
+    feat = transform_features(feat)
 
     # 預測
-    pred_prob = model.predict([seq, feat])[0][0]
+    pred_prob = model.predict([seq, feat], verbose=0)[0][0]
     label = "AI 生成" if pred_prob >= 0.5 else "人類撰寫"
     prob = pred_prob * 100