Java Web学习平台   

import tensorflow as tf

import numpy as np

import pandas as pd

import json

def dataPreProcess(df_data_in):

    # 创建一个新的DataFrame对象，用于存储清洗后的数据

    dfdata_process = pd.DataFrame()

    # Age属性处理----缺失值填充

    # dfdata_in['Aga']表示取出DataFrame中Age这一列

    dfdata_process['Age'] = df_data_in['Age'].fillna(df_data_in['Age'].mean())

    # Fare属性处理,取平均值

    dfdata_process['Fare'] = df_data_in['Fare'].fillna(df_data_in['Fare'].mean())

    # Embarked属性处理 one-hot编码

    embarked_onehot = pd.get_dummies(df_train_raw['Embarked'])

    # 替换为Embarked_C，Embarked_Q，Embarked_S

    embarked_onehot.columns = ['Embarked_' + str(x) for x in embarked_onehot.columns]

    dfdata_process = pd.concat([dfdata_process, embarked_onehot], axis=1)

    # sex属性处理 one-hot编码

    sex_onehot = pd.get_dummies(df_train_raw['Sex'])

    # 替换为Embarked_C，Embarked_Q，Embarked_S

    sex_onehot.columns = ['Embarked_' + str(x) for x in sex_onehot.columns]

    dfdata_process = pd.concat([dfdata_process, sex_onehot], axis=1)

    selected_cols = ['Pclass', 'SibSp', 'Parch']

    dfdata_process = pd.concat([dfdata_process, df_data_in[selected_cols].copy()], axis=1)

    return dfdata_process

TRAIN_DATA_PATH = "D:/train.csv"

TEST_DATA_PATH = "D:/test.csv"

df_train_raw = pd.read_csv(TRAIN_DATA_PATH)

df_test_raw = pd.read_csv(TEST_DATA_PATH)

dftrain_process = dataPreProcess(df_train_raw)

print(dftrain_process)

dftest_process = dataPreProcess(df_test_raw)

models = tf.keras.models

layers = tf.keras.layers

model = models.Sequential()

model.add(layers.Dense(15, activation="relu", input_shape=(10,)))

model.add(layers.Dense(15, activation="relu"))

model.add(layers.Dense(1, activation='sigmoid'))

model.summary()

model.compile(loss='mean_squared_error', optimizer=tf.keras.optimizers.SGD(learning_rate=1e-3), metrics=['AUC'])

dftrain_Serviced = df_train_raw['Survived']

history = model.fit(x=dftrain_process, y=dftrain_Serviced, batch_size=32, epochs=50, validation_split=0.2)

predict_result = model.predict(dftest_process)

predict_result = np.where(predict_result >= 0.5, 1, 0)

pd_result = pd.DataFrame()

pd_result['Survived'] = predict_result.reshape(-1)

pd_result['PassengerId'] = df_test_raw['PassengerId'].astype('Int32')

pd_result = pd_result[0:418]

pd_result.to_csv("predict_result.csv", index=False)

pd_groundtrue = pd.read_csv('D:/gender_submission.csv')

print(np.mean(np.equal(pd_result['Survived'].to_numpy(), pd_groundtrue['Survived'].to_numpy())))

model.save('titanic_model.h5')

json.dump(history.history, open('history.json', 'w'))