学习笔记 Day 42（决策树和随机森林）

信息熵：

划分依据：

信息增益：

API：

 代码实现：

    ssl._create_default_https_context = ssl._create_unverified_context

    # 获取数据
    df = pd.read_csv('https://datahub.csail.mit.edu/download/jander/historic/file/titanic.csv')

    print(df.head())
    # 数据基本处理

    df['age'].fillna(df['age'].mean(), inplace=True, axis=0)

    x = df[['pclass','age','sex']]
    y = df['survived']

    x_train,x_test,y_train,y_test = train_test_split(x,y,test_size=0.2)

    # 数据预处理
    dict = DictVectorizer()

    x_train = dict.fit_transform(x_train.to_dict(orient='records'))
    print(dict.get_feature_names())
    x_test = dict.transform(x_test.to_dict(orient='records'))

    # 机器学习 -- 决策树
    estimator = DecisionTreeClassifier()

    estimator.fit(x_train,y_train)

    y_pre = estimator.predict(x_test)

    print('预测的准确率为：\n',estimator.score(x_test,y_test))
    print('预测的结果为：\n',y_pre)

    # 树的结构保存、
    export_graphviz(estimator,out_file='./tree.dot',feature_names=['年龄', 'pclass=1st', 'pclass=2nd', 'pclass=3rd', '女性', '男性'])

优缺点：

随机森林：

 代码实现：

    # 加验证

    ssl._create_default_https_context = ssl._create_unverified_context

    # 获取数据
    df = pd.read_csv('https://datahub.csail.mit.edu/download/jander/historic/file/titanic.csv')

    print(df.head())
    # 数据基本处理

    df['age'].fillna(df['age'].mean(), inplace=True, axis=0)

    x = df[['pclass','age','sex']]
    y = df['survived']

    x_train,x_test,y_train,y_test = train_test_split(x,y,test_size=0.2)

    # 数据预处理
    dict = DictVectorizer()

    x_train = dict.fit_transform(x_train.to_dict(orient='records'))
    print(dict.get_feature_names())
    x_test = dict.transform(x_test.to_dict(orient='records'))

    # 机器学习 -- 随机森林
    rf = RandomForestClassifier()

    parm = {'n_estimators':[100,200,300,400,500,600],'max_depth':[5,7,12,15,18,21]}

    estimator = GridSearchCV(rf,param_grid=parm,cv=10)

    estimator.fit(x_train,y_train)

    print('准确率为：\n',estimator.score(x_test,y_test))

    print('准确率最高的模型：\n',estimator.best_estimator_)

    print('最好的参数：\n',estimator.best_params_)