[数据科学从零到壹]·泰坦尼克号生存预测（数据读取、处理与建模）​​​​​​​

泰坦尼克号生存预测（数据读取、处理与建模）

• 简介：

本文是泰坦尼克号上的生存概率预测，这是基于Kaggle上的一个经典比赛项目。

数据集：

1.Kaggle泰坦尼克号项目页面下载数据：https://www.kaggle.com/c/titanic

2.网盘地址：https://pan.baidu.com/s/1BfRZdCz6Z1XR6aDXxiHmHA      提取码：jzb3 

• 代码内容

数据读取：

#%%
import tensorflow as tf
import keras
import pandas as pd
import numpy as np

data = pd.read_csv("titanic/train.csv")
print(data.head())
print(data.describe())

数据处理：

#%%
strs = "Survived Pclass Sex Age SibSp Parch Fare Embarked"
clos = strs.split(" ")
print(clos)
#%%
x_datas = data[clos]
print(x_datas.head())
#%%
print(x_datas.isnull().sum())

#%%
x_datas["Age"] = x_datas["Age"].fillna(x_datas["Age"].mean())
x_datas["Embarked"] = x_datas["Embarked"].fillna(x_datas["Embarked"].mode()[0])


#x_datas["Sex"] = pd.get_dummies(x_datas["Sex"])
x_datas = pd.get_dummies(x_datas,columns=["Pclass","Sex","Embarked"])
x_datas["Age"]/=100
x_datas["Fare"]/=100

print(x_datas.isnull().sum())
print(x_datas.head())

#%%
seq = int(0.75*(len(x_datas)))

X ,Y = x_datas.iloc[:,1:],x_datas.iloc[:,0]
X_train,Y_train,X_test,Y_test = X[:seq],Y[:seq],X[seq:],Y[seq:]

模型搭建：

#%%
model = keras.models.Sequential()

model.add(keras.layers.Dense(64,input_dim = 12,activation="relu"))
model.add(keras.layers.Dropout(0.2))
model.add(keras.layers.Dense(16,activation="relu"))
model.add(keras.layers.Dense(2,activation="softmax"))

model.compile(loss="sparse_categorical_crossentropy",optimizer="adam",metrics=["accuracy"])

print(model.summary())

模型训练与评估：

#%%
model.fit(X_train,Y_train,validation_split=0.2,epochs=100,batch_size=50)

#%%
y = model.evaluate(X_test,Y_test)
print("test loss is %f, acc %f"%(y[0],y[1]))
model.save("model_100_1.h5")

• 输出结果：

_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
dense_1 (Dense)              (None, 64)                832
_________________________________________________________________
dropout_1 (Dropout)          (None, 64)                0
_________________________________________________________________
dense_2 (Dense)              (None, 16)                1040
_________________________________________________________________
dense_3 (Dense)              (None, 2)                 34
=================================================================
Total params: 1,906
Trainable params: 1,906
Non-trainable params: 0
_________________________________________________________________
...
Epoch 96/100
534/534 [==============================] - 0s 80us/step - loss: 0.3870 - acc: 0.8277 - val_loss: 0.5083 - val_acc: 0.7612
Epoch 97/100
534/534 [==============================] - 0s 80us/step - loss: 0.3921 - acc: 0.8352 - val_loss: 0.5070 - val_acc: 0.7687
Epoch 98/100
534/534 [==============================] - 0s 82us/step - loss: 0.3940 - acc: 0.8371 - val_loss: 0.5102 - val_acc: 0.7687
Epoch 99/100
534/534 [==============================] - 0s 78us/step - loss: 0.3996 - acc: 0.8277 - val_loss: 0.5106 - val_acc: 0.7687
Epoch 100/100
534/534 [==============================] - 0s 80us/step - loss: 0.3892 - acc: 0.8352 - val_loss: 0.5082 - val_acc: 0.7612
223/223 [==============================] - 0s 63us/step
test loss is 0.389338, acc 0.829596

• 完整代码：

#%%
import tensorflow as tf
import keras
import pandas as pd
import numpy as np

data = pd.read_csv("titanic/train.csv")
print(data.head())
print(data.describe())
#%%
strs = "Survived Pclass Sex Age SibSp Parch Fare Embarked"
clos = strs.split(" ")
print(clos)
#%%
x_datas = data[clos]
print(x_datas.head())
#%%
print(x_datas.isnull().sum())

#%%
x_datas["Age"] = x_datas["Age"].fillna(x_datas["Age"].mean())
x_datas["Embarked"] = x_datas["Embarked"].fillna(x_datas["Embarked"].mode()[0])


#x_datas["Sex"] = pd.get_dummies(x_datas["Sex"])
x_datas = pd.get_dummies(x_datas,columns=["Pclass","Sex","Embarked"])
x_datas["Age"]/=100
x_datas["Fare"]/=100

print(x_datas.isnull().sum())
print(x_datas.head())

#%%
seq = int(0.75*(len(x_datas)))

X ,Y = x_datas.iloc[:,1:],x_datas.iloc[:,0]
X_train,Y_train,X_test,Y_test = X[:seq],Y[:seq],X[seq:],Y[seq:]


#%%
model = keras.models.Sequential()

model.add(keras.layers.Dense(64,input_dim = 12,activation="relu"))
model.add(keras.layers.Dropout(0.2))
model.add(keras.layers.Dense(16,activation="relu"))
model.add(keras.layers.Dense(2,activation="softmax"))

model.compile(loss="sparse_categorical_crossentropy",optimizer="adam",metrics=["accuracy"])

print(model.summary())

#%%
model.fit(X_train,Y_train,validation_split=0.2,epochs=100,batch_size=50)

#%%
y = model.evaluate(X_test,Y_test)
print("test loss is %f, acc %f"%(y[0],y[1]))
model.save("model_100_1.h5")