知识要点

• estimater 有点没理解透

• 数据集是泰坦尼克号人员幸存数据.

• 读取数据：train_df = pd.read_csv('./data/titanic/train.csv')

• 显示数据特征：train_df.info()

• 显示开头部分数据：train_df.head()

• 提取目标特征：y_train = train_df.pop('survived')

• 显示数据分布：train_df.describe()

• 柱状图显示：train_df.age.hist(bins = 20)

• 横向柱状图: train_df.sex.value_counts().plot(kind = 'barh')

• pd.concat([train_df, y_train], axis = 1).groupby('sex').survived.mean().plot(kind = 'barh')  # 根据幸存率查看各类型的均值

• 提取不同特征的统计: train_df.embark_town.value_counts()

• 提取特征: vocab = train_df[categorical_column].unique()

• tf.feature_column.indicator_column(tf.feature_column.categorical_column_with_vocabulary_list(categorical_column, vocab))   # one_hot 编码

• dataset批次设置: dataset = dataset.repeat(epochs).batch(batch_size) 

---

1 导包

from tensorflow import keras
import tensorflow as tf
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

2 数据导入

train_df = pd.read_csv('./data/titanic/train.csv')
eval_df = pd.read_csv('./data/titanic/eval.csv')  # eval 评估   # 数据
print(train_df.info())
print(eval_df.info())

train_df.head()

 3 目标值获取

y_train = train_df.pop('survived')
y_eval = eval_df.pop('survived')

print(train_df.head())
print(eval_df.head())
print(y_train.head())
print(y_eval.head())

4 特征处理

train_df.describe()

# 观察年龄的数据分布
train_df.age.hist(bins = 20)

# 观察男女比例, 性别数量对比
train_df.sex.value_counts().plot(kind = 'barh')

# 仓位对比, 船舱类型
train_df['class'].value_counts().plot(kind = 'barh')

# 看港口人数
train_df['embark_town'].value_counts().plot(kind = 'barh')

pd.concat([train_df, y_train], axis = 1).groupby('sex').survived.mean().plot(kind = 'barh')

train_df.embark_town.value_counts()
'''Southampton    450
Cherbourg      123
Queenstown      53
unknown          1
Name: embark_town, dtype: int64'''

# 区分离散特征和连续特征
categorical_columns = ['sex', 'n_siblings_spouses', 'parch', 'class', 'deck', 'embark_town', 'alone']  # 离散特征
numeric_columns = ['age', 'fare']

# 接受特征
feature_columns = []
for categorical_column in categorical_columns:
    vocab = train_df[categorical_column].unique()  # 取出特征值
    print(vocab)
    # print(tf.feature_column.categorical_column_with_vocabulary_list(categorical_column, vocab))  # 创建vocabulary 的API
    # 将离散特征转换为one_hot形式的编码
    num = tf.feature_column.indicator_column(tf.feature_column.categorical_column_with_vocabulary_list(categorical_column, vocab))
    feature_columns.append(num)

# 数据类型转换
for numeric_column in numeric_columns:
    feature_columns.append(tf.feature_column.numeric_column(numeric_column, dtype = tf.float32))

5 dataset

# 创建生成dataset的方法
def make_dataset(data_df, label_df, epochs = 10, shuffle = True, batch_size = 32):
    dataset = tf.data.Dataset.from_tensor_slices((dict(data_df), label_df))
    if shuffle:
        dataset = dataset.shuffle(10000)  # 打乱, 洗牌
    dataset = dataset.repeat(epochs).batch(batch_size)
    return dataset
train_dataset = make_dataset(train_df, y_train, batch_size = 5)

# baseline_model
import os
output_dir = 'baseline_model'
if not os.path.exists(output_dir):
    os.mkdir(output_dir)
    
baseline_estimator = tf.compat.v1.estimator.BaselineClassifier(model_dir = output_dir, n_classes= 2)
# input_fn要求没有输入参数, 要求返回元组(x, y)或者可以返回(x, y)的dataset
baseline_estimator.train(input_fn = lambda : make_dataset(train_df, y_train, epochs = 100))

# baseline 是随机参数, 所以结果很差
baseline_estimator.evaluate(input_fn = lambda : make_dataset(eval_df, y_eval, epochs = 1,
                                                             shuffle = False, batch_size = 20))

# linear_model
linear_output_dir = 'linear_model'
if not os.path.exists(linear_output_dir):
    os.mkdir(linear_output_dir)
    
linear_estimator = tf.estimator.LinearClassifier(feature_columns = feature_columns,
                                                 model_dir = linear_output_dir)
linear_estimator.train(input_fn = lambda :make_dataset(train_df, y_train, epochs = 100))

# baseline 是随机参数, 所以结果很差
linear_estimator.evaluate(input_fn = lambda : make_dataset(eval_df, y_eval, epochs = 1, shuffle = False,
                                                           batch_size = 20))

dnn_output_dir = './dnn_model'
if not os.path.exists(dnn_output_dir):
    os.mkdir(dnn_output_dir)
    
dnn_estimator = tf.estimator.DNNClassifier(model_dir = dnn_output_dir,  # 存储地址
                                           n_classes= 2,  # 二分类
                                           feature_columns = feature_columns, 
                                           hidden_units = [128, 128],   # 隐藏层
                                           activation_fn = tf.nn.relu,  # 算法
                                           optimizer = 'Adam')  # 损失函数, 优化:optimizer
# dnn_estimator.train(input_fn = lambda : make_dataset(train_df, y_train, epochs = 100))

dnn_estimator.train(input_fn = lambda :make_dataset(train_df, y_train, epochs = 100))

dnn_estimator.evaluate(input_fn = lambda : make_dataset(eval_df, y_eval, epochs = 1,
                                                        shuffle = False, batch_size = 20))