机器学习2：KNN决策树探究泰坦尼克号幸存者问题

  KNN决定树探究泰坦尼克号幸存者问题

1. import pandas as pdfrom sklearn.tree import DecisionTreeClassifier, export_graphvizfrom sklearn.metrics import classification_reportimport graphviz   #决定树可视化

复制代码

1. data = pd.read_csv(r"titanic_data.csv")data.drop("PassengerId",axis = 1,inplace = True)  #删除id这一列

复制代码

1. data

复制代码

  SurvivedPclassSexAge003male22.0111female38.0213female26.0311female35.0403male35.0...............88602male27.088711female19.088803femaleNaN88911male26.089003male32.0 891 rows × 4 columns

1. data.loc[data["Sex"] == "male","Sex"] = 1data.loc[data["Sex"] == "female","Sex"] = 0

复制代码

1. data

复制代码

  SurvivedPclassSexAge003122.0111038.0213026.0311035.0403135.0...............88602127.088711019.0888030NaN88911126.089003132.0 891 rows × 4 columns

1. data.fillna(data["Age"].mean(),inplace = True)  #用均值来填充缺失值

复制代码

1. data

复制代码

  SurvivedPclassSexAge003122.000000111038.000000213026.000000311035.000000403135.000000...............88602127.00000088711019.00000088803029.69911888911126.00000089003132.000000 891 rows × 4 columns

1. Dtc = DecisionTreeClassifier(max_depth = 5,random_state =8)  #构建决定树Dtc.fit(data.iloc[:,1:],data["Survived"])    #模子训练pre = Dtc.predict(data.iloc[:,1:])  #模子预测

复制代码

1. print(classification_report(pre,data["Survived"]))   #肴杂矩阵

复制代码

1.               precision    recall  f1-score   support           0       0.88      0.84      0.86       573           1       0.73      0.79      0.76       318    accuracy                           0.82       891   macro avg       0.81      0.82      0.81       891weighted avg       0.83      0.82      0.82       891

复制代码

1. pre == data["Survived"]   #比力模子预测值与实际值是否一致

复制代码

1. 0       True1       True2       True3       True4       True       ...  886     True887     True888    False889    False890     TrueName: Survived, Length: 891, dtype: bool

复制代码

可视化

1. dot_data = export_graphviz(Dtc,feature_names = ["Pclass","Sex","Age"],class_names="Survive")

复制代码

1. graph  = graphviz.Source(dot_data)graph

复制代码

来源：https://blog.csdn.net/qq_45176548/article/details/112060492
免责声明：如果侵犯了您的权益，请联系站长，我们会及时删除侵权内容，谢谢合作！