十三、(机器学习)-利用CART与LM神经网络进行漏电用户分析

1、背景和分析

本次数据挖掘建模目标如下：
1、归纳出漏窃电用户的关键特征，构建窃漏电用户的识别模型
2、利用实时检测数据，调用窃漏电用户识别模型实现实时的检测

2、代码

import warnings
warnings.filterwarnings('ignore')

import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
%matplotlib inline

from random import shuffle

from keras.layers.core import Activation, Dense
from keras.models import Sequential
from scipy.interpolate import largrange
from sklearn.externals import joblib
from sklearn.metrics import confusion_matrix, toc_curve
from sklearn.tree import DecisionTreeClassifier

使用拉格朗日插值法，插补缺少的数据

input_file = 'E:/1-GitHubCode-book/python_data_analysis_and_mining_action/chapter6/data/missing_data.xls'
out_file = './missing_data_processed.xls'

data = pd.read_excel(input_file, header = None)

def ployinterp_column(s, n, k=5):
# s是series
	y = s[list(range(n-k, n)) + list(range(n+1, k+n+1))]
	
	y = y[y.notnull()]
	return largrange(y.index, list(y))(n)

for i in data.columns:
	for j in range(len(data)):
		if (data[i].isnull()):
			data[i][j] = ployinterp_column(data[i], j)
data.to_excel(out_file, header = None, index = False)

使用CART， 对290个用户进行训练

data_file = 'E:/1-GitHubCode-book/python_data_analysis_and_mining_action/chapter6/data/model.xls'
data = pd.read_excel(data_file)

# 定义混淆矩阵
def cm_plot(y, yp):
	cm = confusion_matrix(y, yp)
	plt.matshow(cm, cmap = plt.cm.Reds)
	plt.colorbar()

	for x in range(len(cm)):
		for y in range(len(cm)):
			plt.annotate(
				cm[x, y],
				xy = (x, y),
				horizontalalignment = 'center',
				verticalalignment = 'center'
			)
		plt.ylabel('True label')
		plt.xlabel('Predicted label')

		return plt
	

data  = data.as_matrix()   # 转换为矩阵
shuffle(data)

p = 0.8
train = data[:int(len(data)*p), :]
test = data[int(len(data)*p), :]

treefile  = './tree.pkl'
tree - DecisionTreeClassifier()
tree.fit(train[:, :3], train[:, 3])

joblib.dump(tree, treefile)

cm_plot(train[:, 3], tree.predict(train[:, :3])).show()

fpr, tpr, thresholds = roc_curve(
	test[:, 3], tree.predict_proba(test[:,  :3])[:, 1], pos_labl=1
)
plt.plot(fpr, tpr, linewidth=2, label='ROC OF CART', color='green')

plt.xlabel('FALSE POSITIVE RATE')
plt.ylabel('TRUE POSITIVE RATE')

plt.ylim(0, 1.05)
plt.xlim(0, 1.05)


plt.show()
print(thresholds)

使用LM神经网络
设定LM神经网络的输入节点为3，输出节点为1，隐层节点数为10， 使用Adam方法求解，对于激活函数，在隐藏层使用Relu(x) = max(x, 0)作为激活函数。

data_file = 'E:/1-GitHubCode-book/python_data_analysis_and_mining_action/chapter6/data/model.xls'
data = pd.read_excel(data_file)
data = data.as_matrix()
shuffle(data)
data

p = 0.8
train = data[:int(len(data)*0.8), :]
test = data[int(len(data)*0.8):, :]

# 构建神经网络
netfile = './net.model'
net = Sequential()
net.add(Dense(10, input_shape = (3, ))) # 添加输入层 3 节点，到隐藏层 10 节点的链接
net.add(Activation('relu'))

net.add(Dense(1, input_shape = (10, ))) #ned.add(Dense(10, 1)) 添加隐藏层10节点到输出层1节点的链接
net.add(Activation('sigmoid'))  # 输出层使用sigmoid函数

net.compile(
	loss = 'binary_crossentropy',
	optimizer = 'adam',
	sample_weight_mode = 'binary'
)

net.fit(train[:, :3], train[:, 3], epoches = 100, batch_size = 1)
net.save_weights(netfile)

predict_result = net.predict_classes(train[:, :3]).reshape(len(data))
result = net.predict(train[:, :3])

cm_plot(train[:, 3], predict_result).show()
predict_result = net.predict(test[:, :3]).reshape(len(test))

fpr, tpr, thresholds = roc_curve(test[:, 3], predict_result, pos_label = 1)

plt.plot(fpr, tpr, linewidth = 2, label = 'ROC OF LM')

plt.xlabel('False positive Rate')
plt.ylabel('True positive Rate')

plt.ylim(0, 1.05)
plt.xlim(0, 1.05)

plt.legend(loc = 4)
plt.show()
print(thresholds)

实验表明，分类准确率为(161 + 58) / (161 + 58 +6 +7) = 94.4%，正常用户被判为漏电用户占正常用户的7 / (161 + 7) = 4.2%, 漏电用户被判为正常用户占正常窃漏电用户的 6 / (6+58)=9.4%

---

望您：
“情深不寿，强极则辱，谦谦君子，温润如玉”。