机器学习——朴素贝叶斯练习题

一、

使用鸢尾花数据训练多项式朴素贝叶斯模型，并评估模型

代码展示： 

from sklearn.datasets import load_iris
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import MultinomialNB

iris = load_iris()

x_train,x_test,y_train,y_test = train_test_split(iris.data,iris.target,test_size=0.3,random_state=42)

model = MultinomialNB()

model.fit(x_train,y_train)

y_pred = model.predict(x_test)
print("预测率：",accuracy_score(y_test,y_pred))

 结果展示：

预测率： 0.9555555555555556

二、

电影评论情感分析

‌项目背景‌：

你在一家电影评论网站工作，需要开发一个情感分析系统来自动分类用户评论是正面还是负面。使用Kaggle上的"IMDB Dataset of 50K Movie Reviews"数据集。

‌数据集链接‌：

IMDB Dataset of 50K Movie Reviews | Kaggle

‌练习题要求‌：

1. 使用Pandas加载并预处理数据
2. 使用Numpy进行特征工程
3. 比较不同朴素贝叶斯变体(高斯、多项式、伯努利)的性能
4. 使用matplotlib绘制性能比较图表

代码展示：

import re
import pandas as pd
from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import GaussianNB, MultinomialNB, BernoulliNB
import matplotlib.pyplot as plt

df = pd.read_csv("./data/IMDB Dataset.csv",encoding="utf-8")
print(df.head())
print(df.shape)

df["sentiment"] = df["sentiment"].map({"positive":1,"negative":0})
# print(df.head())

comment = df["review"]
# print(comment.head())

comment_lists = []
for i in comment:
    # print(i)
    i = i.lower()
    i = re.sub(r'<.*?>', '', i)
    i = re.sub(r'[^a-zA-Z]', ' ', i)

    words = i.split()
    words = [word for word in words if len(word) > 2]

    comment_list = " ".join(words)
    comment_lists.append(comment_list)

# print(comment_list)
df["clean_review"] = comment_lists

transfer = TfidfVectorizer(max_features=5000,ngram_range=(1,2))
x = transfer.fit_transform(df["clean_review"])
y = df["sentiment"]

x_train,x_test,y_train,y_test = train_test_split(x,y,test_size=0.3,random_state=42)

mu_model = MultinomialNB()
mu_model.fit(x_train,y_train)

mu_y_pred = mu_model.predict(x_test)
mu_accuracy = accuracy_score(y_test,mu_y_pred)
print("多项式朴素贝叶斯：",mu_accuracy)

be_model = BernoulliNB()
be_model.fit(x_train,y_train)

be_y_pred = be_model.predict(x_test)
be_accuracy = accuracy_score(y_test,be_y_pred)
print("伯努利朴素贝叶斯：",be_accuracy)

transfer = CountVectorizer(max_features=5000)
x = transfer.fit_transform(comment_lists)

x_dense = x.toarray()

x_train = x_dense[:4000, :]
good_or_bad = df["sentiment"].values
y_train = good_or_bad[:4000]
x_test = x_dense[4000:, :]
y_test = good_or_bad[4000:]

ga_model = GaussianNB()
ga_model.fit(x_train,y_train)
ga_y_pred = ga_model.predict(x_test)
ga_accuracy = accuracy_score(y_test,ga_y_pred)
print("高斯朴素贝叶斯：",ga_accuracy)

models = ['GaussianNB','MultinomialNB','BernoulliNB']
values = [ga_accuracy,mu_accuracy,be_accuracy]

plt.bar(
    models,
    values,
    color=['blue','green','red']
)

plt.title("Comparison of Naive Bayes Variants")
plt.ylabel("Accuracy")
plt.tight_layout()
plt.show()

结果展示：

多项式朴素贝叶斯： 0.8628666666666667
伯努利朴素贝叶斯： 0.8533333333333334
高斯朴素贝叶斯： 0.7214347826086956