diff --git a/.gitignore b/.gitignore
index cc0310c..76e7681 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,8 +3,6 @@
 env/
 venv/
 
-# Model files
-models/
 
 # Environment variables
 .env
diff --git a/models/lightgbm.joblib b/models/lightgbm.joblib
new file mode 100644
index 0000000..76e1104
Binary files /dev/null and b/models/lightgbm.joblib differ
diff --git a/models/lightgbm_classification_report.txt b/models/lightgbm_classification_report.txt
new file mode 100644
index 0000000..85b43b8
--- /dev/null
+++ b/models/lightgbm_classification_report.txt
@@ -0,0 +1,8 @@
+              precision    recall  f1-score   support
+
+         ham       0.98      0.99      0.99       969
+        spam       0.93      0.89      0.91       145
+
+    accuracy                           0.98      1114
+   macro avg       0.96      0.94      0.95      1114
+weighted avg       0.98      0.98      0.98      1114
diff --git a/models/lightgbm_confusion_matrix.png b/models/lightgbm_confusion_matrix.png
new file mode 100644
index 0000000..8505421
Binary files /dev/null and b/models/lightgbm_confusion_matrix.png differ
diff --git a/models/logistic_regression.joblib b/models/logistic_regression.joblib
new file mode 100644
index 0000000..a8c4fc7
Binary files /dev/null and b/models/logistic_regression.joblib differ
diff --git a/models/logistic_regression_classification_report.txt b/models/logistic_regression_classification_report.txt
new file mode 100644
index 0000000..e888fbc
--- /dev/null
+++ b/models/logistic_regression_classification_report.txt
@@ -0,0 +1,8 @@
+              precision    recall  f1-score   support
+
+         ham       0.99      0.99      0.99       969
+        spam       0.92      0.94      0.93       145
+
+    accuracy                           0.98      1114
+   macro avg       0.95      0.96      0.96      1114
+weighted avg       0.98      0.98      0.98      1114
diff --git a/models/logistic_regression_confusion_matrix.png b/models/logistic_regression_confusion_matrix.png
new file mode 100644
index 0000000..84db3f5
Binary files /dev/null and b/models/logistic_regression_confusion_matrix.png differ
diff --git a/src/models/__init__.py b/src/models/__init__.py
new file mode 100644
index 0000000..fae3f00
--- /dev/null
+++ b/src/models/__init__.py
@@ -0,0 +1,42 @@
+from .train import (
+    build_tfidf_vectorizer,
+    build_logistic_regression,
+    build_lightgbm,
+    train_model,
+    save_model,
+    load_model
+)
+from .evaluate import (
+    evaluate_model,
+    print_evaluation_results,
+    plot_confusion_matrix,
+    save_evaluation_results,
+    compare_models
+)
+from .predict import (
+    predict_spam,
+    predict_batch_spam,
+    predict_from_df
+)
+
+__all__ = [
+    # train.py
+    "build_tfidf_vectorizer",
+    "build_logistic_regression",
+    "build_lightgbm",
+    "train_model",
+    "save_model",
+    "load_model",
+    
+    # evaluate.py
+    "evaluate_model",
+    "print_evaluation_results",
+    "plot_confusion_matrix",
+    "save_evaluation_results",
+    "compare_models",
+    
+    # predict.py
+    "predict_spam",
+    "predict_batch_spam",
+    "predict_from_df"
+]
\ No newline at end of file
diff --git a/src/models/evaluate.py b/src/models/evaluate.py
new file mode 100644
index 0000000..fb99e6d
--- /dev/null
+++ b/src/models/evaluate.py
@@ -0,0 +1,159 @@
+from sklearn.metrics import (
+    accuracy_score,
+    precision_score,
+    recall_score,
+    f1_score,
+    classification_report,
+    confusion_matrix
+)
+import matplotlib.pyplot as plt
+import seaborn as sns
+import os
+from typing import Dict, Any, List
+import polars as pl
+
+
+
+def evaluate_model(
+    model,
+    test_df: pl.DataFrame
+) -> Dict[str, Any]:
+    """评估模型"""
+    # 准备测试数据
+    X_test = test_df["clean_text"].to_list()
+    y_test = test_df["label_num"].to_list()
+    
+    # 预测
+    y_pred = model.predict(X_test)
+    y_pred_proba = model.predict_proba(X_test)[:, 1]
+    
+    # 计算指标
+    accuracy = accuracy_score(y_test, y_pred)
+    precision = precision_score(y_test, y_pred, average="macro")
+    recall = recall_score(y_test, y_pred, average="macro")
+    f1 = f1_score(y_test, y_pred, average="macro")
+    
+    # 生成分类报告
+    class_report = classification_report(y_test, y_pred, target_names=["ham", "spam"])
+    
+    # 生成混淆矩阵
+    conf_matrix = confusion_matrix(y_test, y_pred)
+    
+    # 收集结果
+    results = {
+        "accuracy": accuracy,
+        "precision": precision,
+        "recall": recall,
+        "f1": f1,
+        "class_report": class_report,
+        "conf_matrix": conf_matrix,
+        "y_test": y_test,
+        "y_pred": y_pred,
+        "y_pred_proba": y_pred_proba
+    }
+    
+    return results
+
+
+def print_evaluation_results(results: Dict[str, Any]) -> None:
+    """打印评估结果"""
+    print(f"Accuracy: {results['accuracy']:.4f}")
+    print(f"Precision (macro): {results['precision']:.4f}")
+    print(f"Recall (macro): {results['recall']:.4f}")
+    print(f"F1 Score (macro): {results['f1']:.4f}")
+    print("\nClassification Report:")
+    print(results['class_report'])
+
+
+def plot_confusion_matrix(
+    conf_matrix: List[List[int]],
+    save_path: str = None
+) -> None:
+    """绘制混淆矩阵"""
+    plt.figure(figsize=(8, 6))
+    sns.heatmap(
+        conf_matrix,
+        annot=True,
+        fmt="d",
+        cmap="Blues",
+        xticklabels=["ham", "spam"],
+        yticklabels=["ham", "spam"]
+    )
+    plt.xlabel("Predicted Label")
+    plt.ylabel("True Label")
+    plt.title("Confusion Matrix")
+    
+    if save_path:
+        os.makedirs(os.path.dirname(save_path), exist_ok=True)
+        plt.savefig(save_path, dpi=300, bbox_inches="tight")
+        print(f"混淆矩阵已保存到: {save_path}")
+    
+    plt.close()
+
+
+def save_evaluation_results(
+    results: Dict[str, Any],
+    model_name: str,
+    save_dir: str = "./models"
+) -> None:
+    """保存评估结果"""
+    # 确保保存目录存在
+    os.makedirs(save_dir, exist_ok=True)
+    
+    # 保存分类报告
+    report_path = os.path.join(save_dir, f"{model_name}_classification_report.txt")
+    with open(report_path, "w") as f:
+        f.write(results["class_report"])
+    
+    # 保存混淆矩阵
+    conf_matrix_path = os.path.join(save_dir, f"{model_name}_confusion_matrix.png")
+    plot_confusion_matrix(results["conf_matrix"], conf_matrix_path)
+    
+    print(f"评估结果已保存到: {save_dir}")
+
+
+def compare_models(
+    models: Dict[str, Any],
+    test_df: pl.DataFrame
+) -> Dict[str, Dict[str, Any]]:
+    """比较多个模型"""
+    comparison_results = {}
+    
+    for model_name, model in models.items():
+        print(f"\n=== 评估 {model_name} 模型 ===")
+        results = evaluate_model(model, test_df)
+        print_evaluation_results(results)
+        save_evaluation_results(results, model_name)
+        comparison_results[model_name] = results
+    
+    return comparison_results
+
+
+def main():
+    """主函数，用于测试"""
+    from .train import load_model
+    from ..data import load_data, preprocess_data, split_data
+    
+    # 加载和预处理数据
+    df = load_data("../data/spam.csv")
+    processed_df = preprocess_data(df)
+    _, test_df = split_data(processed_df)
+    
+    # 加载模型
+    try:
+        lr_model = load_model("logistic_regression")
+        lgbm_model = load_model("lightgbm")
+        
+        # 比较模型
+        models = {
+            "logistic_regression": lr_model,
+            "lightgbm": lgbm_model
+        }
+        
+        compare_models(models, test_df)
+    except FileNotFoundError:
+        print("模型文件不存在，请先运行train.py训练模型")
+
+
+if __name__ == "__main__":
+    main()
\ No newline at end of file
diff --git a/src/models/predict.py b/src/models/predict.py
new file mode 100644
index 0000000..55eca16
--- /dev/null
+++ b/src/models/predict.py
@@ -0,0 +1,135 @@
+from typing import List, Dict, Any
+import polars as pl
+import warnings
+from sklearn.exceptions import DataConversionWarning
+from .train import load_model
+from ..data import clean_text
+
+# 忽略特征名称不匹配的警告
+warnings.filterwarnings("ignore", message="X does not have valid feature names")
+
+
+def predict_spam(
+    text: str,
+    model_name: str = "lightgbm"
+) -> Dict[str, Any]:
+    """预测单条短信是否为垃圾短信"""
+    # 加载模型
+    model = load_model(model_name)
+    
+    # 清洗文本
+    cleaned_text = clean_text(text)
+    
+    # 预测
+    prediction = model.predict([cleaned_text])[0]
+    probability = model.predict_proba([cleaned_text])[0]
+    
+    # 转换结果
+    label = "spam" if prediction == 1 else "ham"
+    
+    return {
+        "original_text": text,
+        "cleaned_text": cleaned_text,
+        "label": label,
+        "label_num": int(prediction),
+        "probability": {
+            "ham": float(probability[0]),
+            "spam": float(probability[1])
+        }
+    }
+
+
+def predict_batch_spam(
+    texts: List[str],
+    model_name: str = "lightgbm"
+) -> List[Dict[str, Any]]:
+    """批量预测短信是否为垃圾短信"""
+    # 加载模型
+    model = load_model(model_name)
+    
+    # 清洗文本
+    cleaned_texts = [clean_text(text) for text in texts]
+    
+    # 预测
+    predictions = model.predict(cleaned_texts)
+    probabilities = model.predict_proba(cleaned_texts)
+    
+    # 转换结果
+    results = []
+    for i, (text, cleaned_text, prediction, probability) in enumerate(zip(texts, cleaned_texts, predictions, probabilities)):
+        label = "spam" if prediction == 1 else "ham"
+        results.append({
+            "id": i,
+            "original_text": text,
+            "cleaned_text": cleaned_text,
+            "label": label,
+            "label_num": int(prediction),
+            "probability": {
+                "ham": float(probability[0]),
+                "spam": float(probability[1])
+            }
+        })
+    
+    return results
+
+
+def predict_from_df(
+    df: pl.DataFrame,
+    text_column: str = "text",
+    model_name: str = "lightgbm"
+) -> pl.DataFrame:
+    """从DataFrame中预测短信是否为垃圾短信"""
+    # 清洗文本
+    df = df.with_columns(
+        pl.col(text_column).map_elements(clean_text, return_dtype=pl.String).alias("clean_text")
+    )
+    
+    # 准备预测数据
+    texts = df["clean_text"].to_list()
+    
+    # 预测
+    model = load_model(model_name)
+    predictions = model.predict(texts)
+    probabilities = model.predict_proba(texts)
+    
+    # 添加预测结果
+    df = df.with_columns(
+        pl.Series("label_num", predictions).cast(pl.Int64),
+        pl.Series("ham_prob", [p[0] for p in probabilities]).cast(pl.Float64),
+        pl.Series("spam_prob", [p[1] for p in probabilities]).cast(pl.Float64)
+    )
+    
+    # 添加标签
+    df = df.with_columns(
+        pl.when(pl.col("label_num") == 1)
+        .then("spam")
+        .otherwise("ham")
+        .alias("label")
+    )
+    
+    return df
+
+
+def main():
+    """主函数，用于测试"""
+    # 测试单条预测
+    test_text = "Free entry in 2 a wkly comp to win FA Cup final tkts 21st May 2005. Text FA to 87121 to receive entry question(std txt rate)T&C's apply 08452810075over18's"
+    result = predict_spam(test_text)
+    print("单条预测结果:")
+    print(result)
+    
+    # 测试批量预测
+    test_texts = [
+        "Free entry in 2 a wkly comp to win FA Cup final tkts 21st May 2005.",
+        "Ok lar... Joking wif u oni...",
+        "WINNER!! As a valued network customer you have been selected to receivea �900 prize reward!",
+        "I'm gonna be home soon and i don't want to talk about this stuff anymore tonight, k?"
+    ]
+    results = predict_batch_spam(test_texts)
+    print("\n批量预测结果:")
+    for res in results:
+        print(res)
+
+
+if __name__ == "__main__":
+    main()
\ No newline at end of file
diff --git a/src/models/train.py b/src/models/train.py
new file mode 100644
index 0000000..c5f5ddc
--- /dev/null
+++ b/src/models/train.py
@@ -0,0 +1,162 @@
+import os
+import joblib
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.linear_model import LogisticRegression
+from lightgbm import LGBMClassifier
+from sklearn.pipeline import Pipeline
+from sklearn.model_selection import GridSearchCV
+from typing import Tuple, Dict, Any
+import polars as pl
+from dotenv import load_dotenv
+
+# 加载环境变量
+load_dotenv()
+
+# 获取模型保存目录
+MODEL_SAVE_DIR = os.getenv("MODEL_SAVE_DIR", "./models")
+
+
+def build_tfidf_vectorizer() -> TfidfVectorizer:
+    """构建TF-IDF向量化器"""
+    return TfidfVectorizer(
+        stop_words="english",
+        max_features=10000,
+        ngram_range=(1, 2),
+        lowercase=True,
+        token_pattern=r'\b[a-zA-Z0-9_]{1,}\b'
+    )
+
+
+def build_logistic_regression() -> Pipeline:
+    """构建Logistic Regression模型流水线"""
+    pipeline = Pipeline([
+        ("tfidf", build_tfidf_vectorizer()),
+        ("lr", LogisticRegression(
+            random_state=42,
+            max_iter=1000,
+            class_weight="balanced"
+        ))
+    ])
+    return pipeline
+
+
+def build_lightgbm() -> Pipeline:
+    """构建LightGBM模型流水线"""
+    pipeline = Pipeline([
+        ("tfidf", build_tfidf_vectorizer()),
+        ("lgbm", LGBMClassifier(
+            random_state=42,
+            class_weight="balanced",
+            verbose=-1,
+            feature_name="auto"
+        ))
+    ])
+    return pipeline
+
+
+def train_model(
+    train_df: pl.DataFrame,
+    model_type: str = "lightgbm",
+    hyperparam_tune: bool = False
+) -> Tuple[Pipeline, Dict[str, Any]]:
+    """训练模型"""
+    # 准备训练数据
+    X_train = train_df["clean_text"].to_list()
+    y_train = train_df["label_num"].to_list()
+    
+    # 选择模型
+    if model_type == "logistic_regression":
+        pipeline = build_logistic_regression()
+        param_grid = {
+            "lr__C": [0.1, 1.0, 10.0],
+            "lr__solver": ["liblinear", "lbfgs"]
+        }
+    elif model_type == "lightgbm":
+        pipeline = build_lightgbm()
+        param_grid = {
+            "lgbm__n_estimators": [100, 200],
+            "lgbm__learning_rate": [0.1, 0.2],
+            "lgbm__max_depth": [3, 5, 7]
+        }
+    else:
+        raise ValueError(f"不支持的模型类型: {model_type}")
+    
+    # 超参数调优
+    if hyperparam_tune:
+        grid_search = GridSearchCV(
+            pipeline,
+            param_grid,
+            cv=5,
+            scoring="f1_macro",
+            n_jobs=-1
+        )
+        grid_search.fit(X_train, y_train)
+        best_model = grid_search.best_estimator_
+        best_params = grid_search.best_params_
+        print(f"最佳参数: {best_params}")
+    else:
+        best_model = pipeline
+        best_model.fit(X_train, y_train)
+        best_params = {}
+    
+    return best_model, best_params
+
+
+def save_model(model: Pipeline, model_name: str) -> str:
+    """保存模型"""
+    # 确保模型保存目录存在
+    os.makedirs(MODEL_SAVE_DIR, exist_ok=True)
+    
+    # 构建模型文件路径
+    model_path = os.path.join(MODEL_SAVE_DIR, f"{model_name}.joblib")
+    
+    # 保存模型
+    joblib.dump(model, model_path)
+    print(f"模型已保存到: {model_path}")
+    
+    return model_path
+
+
+def load_model(model_name: str) -> Pipeline:
+    """加载模型"""
+    # 构建模型文件路径
+    model_path = os.path.join(MODEL_SAVE_DIR, f"{model_name}.joblib")
+    
+    # 加载模型
+    if not os.path.exists(model_path):
+        raise FileNotFoundError(f"模型文件不存在: {model_path}")
+    
+    model = joblib.load(model_path)
+    print(f"模型已从: {model_path} 加载")
+    
+    return model
+
+
+def main():
+    """主函数，用于测试"""
+    import sys
+    import os
+    
+    # 添加项目根目录到Python路径
+    sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..")))
+    
+    from src.data import load_data, preprocess_data, split_data
+    
+    # 加载和预处理数据
+    df = load_data("./data/spam.csv")
+    processed_df = preprocess_data(df)
+    train_df, test_df = split_data(processed_df)
+    
+    print("训练Logistic Regression模型...")
+    lr_model, lr_params = train_model(train_df, model_type="logistic_regression")
+    save_model(lr_model, "logistic_regression")
+    
+    print("\n训练LightGBM模型...")
+    lgbm_model, lgbm_params = train_model(train_df, model_type="lightgbm")
+    save_model(lgbm_model, "lightgbm")
+    
+    print("\n模型训练完成！")
+
+
+if __name__ == "__main__":
+    main()
\ No newline at end of file