這份資料是我的製程紀錄，defect=1 代表不良。請用程式畫圖： 1. 不良率隨時間的變化 2. 不同班別、機台、材料批號的不良率 3. 每個參數在良品和不良品之間的分布差異 並說明你觀察到的現象，但不要下因果結論。

import pandas as pd from sklearn.ensemble import RandomForestClassifier from sklearn.inspection import permutation_importance from sklearn.metrics import average_precision_score df = pd.read_csv("lots.csv", parse_dates=["time"]).sort_values("time") features = ["temp", "pressure", "speed", "humidity"] cut = int(len(df) * 0.8) # 前 80% 訓練，後 20% 驗證，不能隨機切 train, test = df.iloc[:cut], df.iloc[cut:] model = RandomForestClassifier(n_estimators=300, min_samples_leaf=5, class_weight="balanced", random_state=0) model.fit(train[features], train["defect"]) base = test["defect"].mean() score = average_precision_score(test["defect"], model.predict_proba(test[features])[:, 1]) print(f"不良率基準 {base:.3f}，模型 AP {score:.3f}") imp = permutation_importance(model, test[features], test["defect"], scoring="average_precision", n_repeats=20, random_state=0) print(pd.Series(imp.importances_mean, index=features).sort_values(ascending=False).round(3))