def lrisTrain():
data0 = dataUtil.load_data('../data/data.csv')
data = data0.iloc[:, 1:]
str = [
'number', 'cql', 'vitalCapacity', 'a', 'zxqdcs', 'age', 'BMI', 'label'
]
data = data[str]
# 标准化
# data = dataUtil.standardization(data)
print('*' * 10, 'data(标准化之后)')
print(data.head(10))
# 把数据分为测试数据和验证数据
# 划分x,y
x, y = dataUtil.get_x_y(data)
print('*' * 10, 'x')
print(x)
print('*' * 10, 'y')
print(y)
# 划分训练集测试集
train_data, test_data, train_target, test_target = dataUtil.k_fold(x, y)
print('*' * 10, 'train_data')
print(train_data)
print('*' * 10, 'test_data')
print(test_data)
print('*' * 10, 'train_target')
print(train_target)
print('*' * 10, 'test_target')
print(test_target)
# Model(建模)-引入决策树
# 决策树
# clf = tree.DecisionTreeClassifier(criterion="entropy")
# AdaBoost
clf = tree.DecisionTreeRegressor()
# clf = tree.AdaBoostClassifier(
# base_estimator=tree.DecisionTreeClassifier(max_depth=5, min_samples_split=30, min_samples_leaf=5),
# n_estimators=10, learning_rate=0.2)
# 训练集进行训练
clf.fit(train_data, train_target)
# # 进行预测
y_pred = clf.predict(test_data)
# # 法一：通过准确率进行验证
print(metrics.accuracy_score(y_true=test_target, y_pred=y_pred))
# 画图方法1-生成dot文件
# with open('treeone.dot', 'w') as f:
# dot_data = tree.export_graphviz(clf, out_file=None)
# f.write(dot_data)
# 画图方法2-生成pdf文件
# dot_data = tree.export_graphviz(clf, out_file=None, feature_names=clf.feature_importances_,
# filled=True, rounded=True, special_characters=True)
dot_data = tree.export_graphviz(clf,
out_file=None,
filled=True,
rounded=True,
special_characters=True)
graph = pydotplus.graph_from_dot_data(dot_data)
# # 保存图像到pdf文件
graph.write_pdf("treetwo.pdf")
def draw_RF_DT():
plt.rcParams['font.sans-serif'] = ['SimHei'] ## 中文黑体
# # data
RF = dataUtil.load_data('../data/RF.csv').iloc[:, 1:]
DT = dataUtil.load_data('../data/DT.csv').iloc[:, 1:]
RF_ACC = RF.iloc[:, 0].values
RF_Precision = RF.iloc[:, 1].values
RF_Recall = RF.iloc[:, 2].values
DT_ACC = DT['ACC'].values
DT_Precision = DT.iloc[:, 1].values
DT_Recall = DT.iloc[:, 2].values
x = [1, 2, 3, 4, 5, 6]
lw = 1
fig = plt.figure()
ax1 = fig.add_subplot(1, 1, 1)
ax1.plot(x, RF_ACC, color='r', lw=lw, linestyle='-', label='RF_ACC')
ax1.plot(x, DT_ACC, color='r', lw=lw, linestyle='--', label='DT_ACC')
ax1.plot(x,
RF_Precision,
color='g',
lw=lw,
linestyle='-',
label='RF_Precision')
ax1.plot(x,
DT_Precision,
color='g',
lw=lw,
linestyle='--',
label='DT_Precision')
ax1.plot(x, RF_Recall, color='b', lw=lw, linestyle='-', label='RF_Recall')
ax1.plot(x, DT_Recall, color='b', lw=lw, linestyle='--', label='DT_Recall')
# ax1.set_xlim([9, 7, 5, 3, 2, 1])
ax1.set_xlabel('子集编号')
ax1.set_ylabel('指标')
# ax1.set_title('P-R曲线')
plt.legend(loc="lower right")
plt.savefig('../pic/RF_DT.png', dpi=400, bbox_inches='tight')
plt.show()
# encoding=utf-8
"""
@Time : 2020/6/20 15:35
@Author : LiuYanZhe
@File : GA.py
@Software: PyCharm
@Description: 使用遗传算法优化神经网络
"""
import matplotlib.pyplot as plt
from sko.GA import GA
from math_model.py import ANN
import numpy as np
import pandas as pd
from math_model.util import dataUtil
data = dataUtil.load_data('../data/data_22.csv')
print('data shape:', data.shape)
# data = data.iloc[:, 2:]
# 标准化
# data = dataUtil.standardization(data)
input_num = 14
# 划分x,y
x, y = dataUtil.get_x_y(data)
print('x:', x.shape)
print('y:', y.shape)
# 正则化
# x = dataUtil.normalization(x)
# 标准化
x = dataUtil.standardization2(x)
# 归一化
# x = dataUtil.scale(x)
def get_probability(index, data):
data1 = data[index]
d_0 = data1[data1['label'] == 0]
d_1 = data1[data1['label'] == 1]
return len(d_1) / (len(d_1) + len(d_0))
# 返回类中的阴性点
def get_yin(index, data):
data1 = data[index]
d_0 = data1[data1['label'] == 0]
num = d_0['number']
return num
data0 = dataUtil.load_data('../data/data.csv')
# 标准化
# data0 = dataUtil.standardization(data0)
data = data0.iloc[:, 1:data0.shape[1] - 1]
# data0 = data0.iloc[:, 1:]
# data0 = data0.drop(['wzqdcsd', 'wzqdcsC', 'wzqdcsA', 'wzqdcsB'], axis=1)
# data0 = data0.drop(['wzqdcsC', 'wzqdcsA', 'wzqdcsB'], axis=1)
# 归一化
# data = dataUtil.scale(data)
best_j = [] # 类别个数
best_sc = [] # 比例
best_i = [] # 第几类
num_list = set() # 创建集合，存储num，不重复
print(data)
for j in range(2, 100):
print('*' * 20, j)
# encoding=utf-8
"""
@Time : 2020/6/21 21:34
@Author : LiuYanZhe
@File : set_nan.py
@Software: PyCharm
@Description: 填充缺失值
"""
from sklearn.linear_model import Ridge, Lasso
from math_model.util import dataUtil
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
# 读取数据
data0 = dataUtil.load_data('../data/data_nan.csv')
data_nan, data_no_nan, y_train = dataUtil.get_nan_nonan(data0)
# 划分数据
x_train = data_no_nan.iloc[:, 6:data_no_nan.shape[1] - 1]
x_pre = data_nan.iloc[:, 6:data_no_nan.shape[1] - 1]
print(x_train)
print(x_pre)
print(y_train)
# clf = Ridge(alpha=.5)
clf = Ridge()
clf.fit(x_train, y_train)
print(clf.coef_) # 相关系数
print(clf.intercept_) # 截距
y_pre = clf.predict(x_pre)
print('*' * 10, 'pre')
print(y_pre)