目录

1逻辑回归介绍

2 泰坦尼克号数据集介绍及预处理

2.1数据集介绍

2.2数据预处理

3用tensorflow搭建网络进行训练和预测

3.1搭建训练网络

3.2进行训练

3.3测试集进行预测，并上传kaggle查看正确率

4可视化输出结果

1逻辑回归介绍

其实我们可以把逻辑回归当成只有一层的神经网络。关于逻辑回归的具体知识可以参考我的上一边博客，地址：/postedit/82929291。使用交叉熵损失函数，下面两种形式都可以：

2 泰坦尼克号数据集介绍及预处理

2.1数据集介绍

数据集分为：train.csv和 test.csv

2.2数据预处理

用pandas和numpy进行处理数据集。把 PassengerId，Name，Ticket列去掉。把字符串类型变成数值型，缺失的值进行填充。把训练数据分成输入特征x和标签y。在jupyter notebook编写代码，训练数据预处理，具体代码如下：

import numpy as npimport pandas as pddata = pd.read_csv('train.csv')data = data[['Survived', 'Pclass', 'Sex', 'Age', 'SibSp','Parch', 'Fare', 'Cabin', 'Embarked']] #筛选需要的列data['Age'] = data['Age'].fillna(data['Age'].mean()) #平均值填充data['Cabin'] = pd.factorize(data.Cabin)[0] #数值化data.fillna(0,inplace=True) #剩余的缺失值用0填充data['Sex'] = [1 if x == 'male' else 0 for x in data.Sex] #男变为1女变为0data['p1'] = np.array(data['Pclass'] == 1).astype(np.int32)#船舱等级变为三列data['p2'] = np.array(data['Pclass'] == 2).astype(np.int32)data['p3'] = np.array(data['Pclass'] == 3).astype(np.int32)del data['Pclass']#删除此列data['e1'] = np.array(data['Embarked'] == 'S').astype(np.int32)#也变为三列data['e2'] = np.array(data['Embarked'] == 'C').astype(np.int32)data['e3'] = np.array(data['Embarked'] == 'Q').astype(np.int32)del data['Embarked']data_train = data[[ 'Sex', 'Age', 'SibSp', 'Parch', 'Fare', 'Cabin', 'p1', 'p2','p3', 'e1', 'e2', 'e3']]data_target = data['Survived'].values.reshape(len(data),1)

测试数据预处理，代码如下：

data_test = pd.read_csv('test.csv')data_test = data_test[['Pclass', 'Sex', 'Age', 'SibSp','Parch', 'Fare', 'Cabin', 'Embarked']]data_test['Age'] = data_test['Age'].fillna(data_test['Age'].mean())data_test['Cabin'] = pd.factorize(data_test.Cabin)[0]#数值化data_test.fillna(0,inplace = True)data_test['Sex'] = [1 if x == 'male' else 0 for x in data_test.Sex]data_test['p1'] = np.array(data_test['Pclass'] == 1).astype(np.int32)data_test['p2'] = np.array(data_test['Pclass'] == 2).astype(np.int32)data_test['p3'] = np.array(data_test['Pclass'] == 3).astype(np.int32)data_test['e1'] = np.array(data_test['Embarked'] == 'S').astype(np.int32)data_test['e2'] = np.array(data_test['Embarked'] == 'C').astype(np.int32)data_test['e3'] = np.array(data_test['Embarked'] == 'Q').astype(np.int32)del data_test['Pclass']del data_test['Embarked']

3用tensorflow搭建网络进行训练和预测

3.1搭建训练网络

import tensorflow as tfx = tf.placeholder(dtype='float',shape=[None,12])y = tf.placeholder(dtype='float',shape=[None,1])#前向传播过程weight = tf.Variable(tf.random_normal([12,1]))bias = tf.Variable(tf.random_normal([1]))output = tf.matmul(x,weight) + biaspred = tf.cast(tf.sigmoid(output) > 0.5,tf.float32)#预测结果大于0.5值设为1，否则为0loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels=y,logits=output))#交叉熵损失函数train_step = tf.train.GradientDescentOptimizer(0.0003).minimize(loss)#梯度下降法训练accuracy = tf.reduce_mean(tf.cast(tf.equal(pred,y),tf.float32))

3.2进行训练

sess = tf.Session() #开启会话sess.run(tf.global_variables_initializer())#初始化所有变量loss_train = []train_acc = []for i in range(25000):#index = np.random.permutation(len(data_target))#打乱数据顺序，防止过拟合#data_train = data_train[index]#data_target = data_target[index]for n in range(len(data_target)//100 + 1):batch_xs = data_train[n*100:n*100+100]batch_ys = data_target[n*100:n*100+100]sess.run(train_step,feed_dict={x: batch_xs,y: batch_ys})if i % 5000 ==0:loss_temp = sess.run(loss,feed_dict={x: batch_xs,y: batch_ys})loss_train.append(loss_temp)train_acc_temp = sess.run(accuracy,feed_dict={x: batch_xs,y: batch_ys})train_acc.append(train_acc_temp)print(loss_temp,train_acc_temp)

输出结果如下：

3.3测试集进行预测，并上传kaggle查看正确率

predictions = sess.run(pred,feed_dict={x:data_test})data_test = pd.read_csv('test.csv')predictions = predictions.flatten()#二维数组变成一维数组submission = pd.DataFrame({"PassengerId": data_test["PassengerId"],"Survived": predictions})submission.to_csv("titanic-submission.csv", index=False)

将保存的 titanic-submission.csv文件上传到kaggle网站，进行查看测试集的正确率。若显示正确率为零，可能就文件的格式不对，您需要打开文件然后另存为，选择utf-8编码的格式即可。下面是我提交的结果：

4可视化输出结果

import matplotlib.pyplot as pltplt.plot(loss_train,'k-',label = 'loss_train')plt.title('train_loss')plt.legend()plt.show()

可视化结果：

完整的代码和数据集在github：/xiaooudong/tensorflow_titanic_logistic_regression