【TensorFlow 分类案例】银行客户流失预测

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import datetime, os

from sklearn.preprocessing import StandardScaler, LabelEncoder
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix

import tensorflow as tf
from tensorflow import keras

import warnings
warnings.filterwarnings("ignore")

1️⃣数据加载

2️⃣数据可视化分析

3️⃣特征工程

🎯

🎯 

🎯

🎯

4️⃣网络搭建与训练

model = keras.models.Sequential()

# 第一层 ： 128个node
model.add(keras.layers.Dense(128, activation='relu', input_dim=X_train.shape[1])) # input_dim与特征的数量一致

# 第二层 ： 64个node
model.add(keras.layers.Dense(64, activation='relu'))

# 第三层 ： 32个node
model.add(keras.layers.Dense(32, activation='relu'))

# 第四层 ： 1个输出
model.add(keras.layers.Dense(1, activation='sigmoid')) # sigmoid的输出0或1，对应两种状态

🍒模型编译

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

🍒日志保存

logdir = os.path.join("log", datetime.datetime.now().strftime("%Y%m%d-%H%M%S"))

tensorboard_callback = tf.keras.callbacks.TensorBoard(logdir, histogram_freq=1)

🍒模型训练

output = model.fit(X_train, y_train, batch_size=32, epochs=30, validation_data=(X_test, y_test),callbacks=[tensorboard_callback])

# 显示验证accuracy和loss
plt.figure(figsize=(10,8))
plt.plot(output.history['val_accuracy'], label='val_acc')
plt.plot(output.history['val_loss'], label='val_loss')
plt.legend()
plt.title("Val Result")
plt.show()

5️⃣tensorboard 结果可视化

%load_ext tensorboard

%tensorboard --logdir log --port 6006  # 魔法操作

 

6️⃣模型评估和预测

y_pred = model.predict(X_test) 
y_pred = np.argmax(y_pred,axis=1)
y_pred[:10] # 预测标签类别