机器学习领域：泰坦尼克号生还预测 数据集及完整代码

# -*- coding: utf-8 -*- """ Created on Sat Apr 7 09:50:12 2018 @author: Python Learner """ import pandas as pd import numpy as np import re from sklearn.preprocessing import StandardScaler from sklearn.svm import SVC from sklearn.neural_network import MLPClassifier as MLP from sklearn.ensemble import RandomForestClassifier as RF from sklearn.ensemble import GradientBoostingClassifier as GB from sklearn.ensemble import VotingClassifier train_file = r'datasets/train.csv' test_file = r'datasets/test.csv' data = pd.read_csv(train_file,index_col='PassengerId') data_sub = pd.read_csv(test_file,index_col='PassengerId') data_copy = data.copy() del data_copy['Survived'] data_all = pd.concat([data_copy,data_sub]) #数据合并 #根据Title填充Age空值 def get_title(name): title_search = re.search("([A-Za-z]+)\.",name) if title_search: return title_search.group(1) return "" data_all['Title'] = data_all.Name.apply(get_title) for title in data_all[data_all.Age.isnull()].Title.unique(): title_age_mean = data_all[data_all.Title == title].Age.mean() data_all.loc[data_all.Age.isnull()*data_all.Title == title,'Age'] = \ title_age_mean #填充Fare与Embark空值 Fare_mean = data_all[data_all.Pclass == 3].Fare.mean() #计算均值 Embarked_mode = data_all.Embarked.mode()[0] #计算众数 data_all.Embarked=data_all.Embarked.fillna(Embarked_mode) #众数填充 data_all.Fare=data_all.Fare.fillna(Fare_mean) #均值填充 #年龄离散化 bins=[0,14,30,45,60,80] cats=pd.cut(data_all.Age.as_matrix(),bins) data_all.Age=cats.codes #Fare归一化 scaler=StandardScaler() data_all.Fare=scaler.fit_transform(data_all.Fare.values.reshape(-1,1)) #data_all["Fare"] = data_all["Fare"].map(lambda i: np.log(i) if i > 0 else 0) data_all['FamilySize'] = data_all.Parch + data_all.SibSp data_all['Cabin_null'] = np.array(data_all.Cabin.isnull()).astype(np.int32) data_all['Cabin_nnull'] = np.array(data_all.Cabin.notnull()).astype(np.int32) Sex_dummies = pd.get_dummies(data_all.Sex, prefix= 'Sex') Pclass_dummies = pd.get_dummies(data_all.Pclass,prefix= 'Pclass') Embarked_dummies = pd.get_dummies(data_all.Embarked,prefix= 'Embarked') data_all = pd.concat([data_all, Sex_dummies, Pclass_dummies, Embarked_dummies], axis=1) feature = [ 'Age','Fare','FamilySize', 'Cabin_null','Cabin_nnull','Sex_female','Sex_male', 'Pclass_1','Pclass_2','Pclass_3', 'Embarked_C','Embarked_Q','Embarked_S'] X = data_all.loc[data.index][feature] y = data.Survived from sklearn.model_selection import GridSearchCV, StratifiedKFold kfold = StratifiedKFold(n_splits=10) clf_RF = RF() rf_param_grid = {"max_depth": [None], "max_features": [1, 3, 10], "min_samples_split": [2, 3, 10], "min_samples_leaf": [1, 3, 10], "bootstrap": [False], "n_estimators" :[100,300,500], "criterion": ["gini"]} gsRF = GridSearchCV(clf_RF,param_grid = rf_param_grid, cv=kfold, scoring="accuracy", n_jobs= 4, verbose = 1) gsRF.fit(X,y) rf_best = gsRF.best_estimator_ clf_SVC = SVC(probability=True) svc_param_grid = {'kernel': ['rbf'], 'gamma': [ 0.001, 0.01, 0.1, 1], 'C': [1, 10, 50, 100,200,300, 1000]} gsSVC = GridSearchCV(clf_SVC,param_grid = svc_param_grid, cv=kfold, scoring="accuracy", n_jobs= 4, verbose = 1) gsSVC.fit(X,y) svm_best = gsSVC.best_estimator_ clf_GB = GB() gb_param_grid = {'loss' : ['deviance'], 'n_estimators' : [100,300,500], 'learning_rate': [0.1, 0.05, 0.01], 'max_depth': [4, 8], 'min_samples_leaf': [100,150], 'max_features': [0.3, 0.1]} gsGB = GridSearchCV(clf_GB,param_grid = gb_param_grid, cv=kfold, scoring="accuracy", n_jobs= 4, verbose = 1) gsGB.fit(X,y) gb_best = gsGB.best_estimator_ clf_MLP = MLP() mlp_param_grid = {'hidden_layer_sizes' : [100,200,300,400,500], 'activation' : ['relu'], 'solver' : ['adam'], 'learning_rate_init': [0.01, 0.001], 'max_iter': [5000]} gsMLP = GridSearchCV(clf_MLP,param_grid = mlp_param_grid, cv=kfold, scoring="accuracy", n_jobs= 4, verbose = 1) gsMLP.fit(X,y) mlp_best = gsMLP.best_estimator_ votingC = VotingClassifier(estimators=[('clf_GB', gb_best), ('clf_RF', rf_best),('clf_SVC', svm_best), ('clf_MLP',mlp_best)],voting='soft', n_jobs=4) votingC = votingC.fit(X, y) X_sub = data_all.loc[data_sub.index][feature] #提取测试数据特征 y_sub = votingC.predict(X_sub) #使用模型预测数据标签 result = pd.DataFrame({'PassengerId':data_sub.index,'Survived':y_sub}) result.to_csv(r'D:\[DataSet]\1_Titanic\submission.csv', index=False)