NLP09-加强1-对比SVM

支持向量机（SVM）

（一）导入 SVM 相关库

（二）修改模型初始化

（三）比较

朴素贝叶斯分类器

SVM分类器

支持向量机（SVM）

代码修改基于NLP09-朴素贝叶斯问句分类（3/3）

（一）导入 SVM 相关库

from sklearn.svm import SVC  # 导入 SVM

（二）修改模型初始化

    # 模型训练def train_model(self):self.to_vect()# 使用 SVM 替换朴素贝叶斯svm_model = SVC(kernel='linear', C=1.0)  # 线性核函数，C 是正则化参数svm_model.fit(self.train_vec, self.train_y)self.model = svm_model

详细解释SVM

参见 机器学习——支持向量机（SVM）

# 使用 SVM 替换朴素贝叶斯svm_model = SVC(kernel='linear', C=1.0)  # 线性核函数，C 是正则化参数

（三）比较

性能评估指标主要是：准确性、精确率、召回率、F1-Score

朴素贝叶斯分类器

为了进行性能评估，我们需要使用 train_test_split 来分割数据集，并使用 sklearn.metrics 来计算准确性、精确率、召回率和 F1-Score。下面是修改后的完整代码，包含了数据集划分和各项评估指标的计算：

import os.path
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from common import constant
from ch import data_loader, nlp_utilclass QuestionClassify:def __init__(self):self.train_x = Noneself.train_y = Noneself.tfidf_vec = Noneself.train_vec = Noneself.model = Noneself.question_category_dict = None# 文本向量化def to_vect(self):if self.tfidf_vec is None:# 加载训练数据self.train_x, self.train_y = data_loader.load_train_data()# 初始化一个Tfidfself.tfidf_vec = TfidfVectorizer()# 确保 self.train_x 是字符串列表if isinstance(self.train_x[0], list):self.train_x = [" ".join(doc) for doc in self.train_x]self.train_vec = self.tfidf_vec.fit_transform(self.train_x).toarray()# 模型训练def train_model(self):self.to_vect()# 使用 train_test_split 划分训练集和测试集X_train, X_test, y_train, y_test = train_test_split(self.train_vec, self.train_y, test_size=0.2, random_state=42)# 使用朴素贝叶斯模型nb_model = MultinomialNB(alpha=0.01)nb_model.fit(X_train, y_train)  # 训练模型self.model = nb_model# 预测并计算评估指标y_pred = self.model.predict(X_test)# 计算并打印评估指标accuracy = accuracy_score(y_test, y_pred)precision = precision_score(y_test, y_pred, average='weighted')recall = recall_score(y_test, y_pred, average='weighted')f1 = f1_score(y_test, y_pred, average='weighted')print(f"Accuracy: {accuracy:.4f}")print(f"Precision: {precision:.4f}")print(f"Recall: {recall:.4f}")print(f"F1-Score: {f1:.4f}")# 模型预测def predict(self, question):# 词性标注做电影相关实体的抽取question_cut = nlp_util.movie_pos(question)# 原问句列表（刘德华演过哪些电影）question_src_list = []# 转换后的问句（nr演过哪些电影）question_pos_list = []for item in question_cut:question_src_list.append(item.word)if item.flag in ['nr', 'nm', 'nnt']:question_pos_list.append(item.flag)else:question_pos_list.append(item.word)question_pos_text = [" ".join(question_pos_list)]# 文本向量化question_vect = self.tfidf_vec.transform(question_pos_text).toarray()# 输入模型进行预测，得到结果predict = self.model.predict(question_vect)[0]return predictdef init_question_category_dict(self):# 读取问题（类别-描述）映射文件question_category_path = os.path.join(constant.DATA_DIR, "question_classification.txt")with open(question_category_path, "r", encoding="utf-8") as file:question_category_list = file.readlines()self.question_category_dict = {}for category_item in question_category_list:category_id, category_desc = category_item.strip().split(":")self.question_category_dict[int(category_id)] = category_descdef get_question_desc(self, category):if self.question_category_dict is None:self.init_question_category_dict()return self.question_category_dict[category]if __name__ == "__main__":classify = QuestionClassify()classify.train_model()  # 训练模型并打印评估指标result = classify.predict("刘德华和成龙合作演过哪些电影呢？&&")print(classify.get_question_desc(result))print(result)

修改代码解析：

# 使用 train_test_split 划分训练集和测试集
X_train, X_test, y_train, y_test = train_test_split(self.train_vec, self.train_y, test_size=0.2, random_state=42)
详见 NLP06-Scikit-Learn 机器学习库（鸢尾花为例）的数据集拆分部分。

        # 预测并计算评估指标y_pred = self.model.predict(X_test)# 计算并打印评估指标accuracy = accuracy_score(y_test, y_pred)precision = precision_score(y_test, y_pred, average='weighted')recall = recall_score(y_test, y_pred, average='weighted')f1 = f1_score(y_test, y_pred, average='weighted')
这几个指标是常用的分类模型评估指标。

（1）准确率（Accuracy）

（2）精确率（Precision）

（3）召回率（Recall）

（4） F1-Score

输出结果：

SVM分类器

import os.path
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.svm import SVC  # 导入 SVM
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from common import constant
from ch import data_loader, nlp_utilclass QuestionClassify:def __init__(self):self.train_x = Noneself.train_y = Noneself.tfidf_vec = Noneself.train_vec = Noneself.model = Noneself.question_category_dict = None# 文本向量化def to_vect(self):if self.tfidf_vec is None:# 加载训练数据self.train_x, self.train_y = data_loader.load_train_data()# 初始化一个Tfidfself.tfidf_vec = TfidfVectorizer()# 确保 self.train_x 是字符串列表if isinstance(self.train_x[0], list):self.train_x = [" ".join(doc) for doc in self.train_x]self.train_vec = self.tfidf_vec.fit_transform(self.train_x).toarray()# 模型训练def train_model(self):self.to_vect()# 使用 train_test_split 划分训练集和测试集X_train, X_test, y_train, y_test = train_test_split(self.train_vec, self.train_y, test_size=0.2, random_state=42)# 使用 SVM（支持向量机）替换朴素贝叶斯svm_model = SVC(kernel='linear', C=1.0)  # 线性核函数，C 是正则化参数svm_model.fit(X_train, y_train)  # 训练模型self.model = svm_model# 预测并计算评估指标y_pred = self.model.predict(X_test)# 计算并打印评估指标accuracy = accuracy_score(y_test, y_pred)precision = precision_score(y_test, y_pred, average='weighted', zero_division=0)recall = recall_score(y_test, y_pred, average='weighted', zero_division=0)f1 = f1_score(y_test, y_pred, average='weighted')print(f"Accuracy: {accuracy:.4f}")print(f"Precision: {precision:.4f}")print(f"Recall: {recall:.4f}")print(f"F1-Score: {f1:.4f}")# 模型预测def predict(self, question):# 词性标注做电影相关实体的抽取question_cut = nlp_util.movie_pos(question)# 原问句列表（刘德华演过哪些电影）question_src_list = []# 转换后的问句（nr演过哪些电影）question_pos_list = []for item in question_cut:question_src_list.append(item.word)if item.flag in ['nr', 'nm', 'nnt']:question_pos_list.append(item.flag)else:question_pos_list.append(item.word)question_pos_text = [" ".join(question_pos_list)]# 文本向量化question_vect = self.tfidf_vec.transform(question_pos_text).toarray()# 输入模型进行预测，得到结果predict = self.model.predict(question_vect)[0]return predictdef init_question_category_dict(self):# 读取问题（类别-描述）映射文件question_category_path = os.path.join(constant.DATA_DIR, "question_classification.txt")with open(question_category_path, "r", encoding="utf-8") as file:question_category_list = file.readlines()self.question_category_dict = {}for category_item in question_category_list:category_id, category_desc = category_item.strip().split(":")self.question_category_dict[int(category_id)] = category_descdef get_question_desc(self, category):if self.question_category_dict is None:self.init_question_category_dict()return self.question_category_dict[category]if __name__ == "__main__":classify = QuestionClassify()classify.train_model()  # 训练模型并打印评估指标result = classify.predict("刘德华和成龙合作演过哪些电影呢？&&")print(classify.get_question_desc(result))print(result)

输出结果：

分析：