from sklearn.metrics import accuracy_score
from sklearn.model_selection import StratifiedKFold
import xgboost as xgb
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
import sklearn.preprocessing as preprocessing
# 导入数据
col_names = ["ID","K1K2驱动信号","电子锁驱动信号","急停信号","门禁信号","THDV-M","THDI-M","label"]
data = pd.read_csv("data_train.csv",names=col_names)
dataset_X = data[["K1K2驱动信号","电子锁驱动信号","急停信号","门禁信号","THDV-M","THDI-M"]]#二维数组的格式
dataset_Y = data[["label"]].as_matrix()
dataset_Y=np.array(dataset_Y).reshape(len(dataset_Y))
data.fillna(0)
#数据预处理
scaler = preprocessing.StandardScaler()
lists = ["K1K2驱动信号", "电子锁驱动信号", "急停信号", "门禁信号", "THDV-M", "THDI-M", ]
for list in lists:
data[list] = scaler.fit_transform(data[[list]])
#划分数据集
x_train, x_test, y_train, y_test = train_test_split(dataset_X, dataset_Y,test_size=0.2,random_state=1)
xgb_val = xgb.DMatrix(x_test, label=y_test)
xgb_train = xgb.DMatrix(x_train, label=y_train)
xgb_test = xgb.DMatrix(x_test)
#参数
params = {
'booster': 'gbtree',
# 'objective': 'multi:softmax', # 多分类的问题、
# 'objective': 'multi:softprob', # 多分类概率
'objective': 'binary:logistic',
'eval_metric': 'logloss',
# 'num_class': 9, # 类别数，与 multisoftmax 并用
'gamma': 0.1, # 用于控制是否后剪枝的参数,越大越保守，一般0.1、0.2这样子。
'max_depth': 8, # 构建树的深度，越大越容易过拟合 important
'alpha': 0, # L1正则化系数
'lambda': 11, # 控制模型复杂度的权重值的L2正则化项参数，参数越大，模型越不容易过拟合。
'subsample': 0.7, # 随机采样训练样本
'colsample_bytree': 0.5, # 生成树时进行的列采样
# 'min_child_weight': 2,
'min_child_weight': 3,
# 这个参数默认是 1，是每个叶子里面 h 的和至少是多少，对正负样本不均衡时的 0-1 分类而言
# ，假设 h 在 0.01 附近，min_child_weight 为 1 意味着叶子节点中最少需要包含 100 个样本。
# 这个参数非常影响结果，控制叶子节点中二阶导的和的最小值，该参数值越小，越容易 overfitting。
'silent': 0, # 设置成1则没有运行信息输出，最好是设置为0.
'eta': 0.2, # 如同学习率
'seed': 100,
'nthread': -1, # cpu 线程数
'missing': 1,
'scale_pos_weight': (np.sum(y_train == 0) / np.sum(y_train == 1)),
# 用来处理正负样本不均衡的问题,通常取：sum(negative cases) / sum(positive cases)
# 'eval_metric': 'auc'
}
params['is_unbalance']='true'
params['metric'] = 'auc'
num_rounds = 4000 # 迭代次数
watchlist = [(xgb_train, 'train'), (xgb_val, 'val')]
# 交叉验证
result = xgb.cv(params, xgb_train, num_boost_round=10, nfold=4, early_stopping_rounds=200,
verbose_eval=True,
folds=StratifiedKFold(n_splits=6).split(x_train, y_train))
# early_stopping_rounds 当设置的迭代次数较大时，early_stopping_rounds
可在一定的迭代次数内准确率没有提升就停止训练
# 训练模型并保存
model = xgb.train(params, xgb_train, num_rounds,watchlist)
# model.save_model('../data/model/xgb.model') # 用于存储训练出的模型
preds = model.predict(xgb_test)
print(preds)
for i in range(len(preds)):
if preds[i]>0.51:
preds[i]=1
else:
preds[i]=0
y_predict = [int(item) for item in preds]
print(len(y_predict))
acc=accuracy_score(y_predict,y_test)
print(acc)
#def read_data():
col_names = ["ID", "K1K2驱动信号", "电子锁驱动信号", "急停信号", "门禁信号", "THDV-M", "THDI-M"]
data = pd.read_csv("data_test.csv", names=col_names)
# print([图片]data.info())
data = data.fillna(0)
return data[["K1K2驱动信号", "电子锁驱动信号", "急停信号", "门禁信号", "THDV-M", "THDI-M"]],data["ID"]
#
subdata ,Id = read_data()
xgb_test1 = xgb.DMatrix(subdata)
y_predict = model.predict(xgb_test1)
print(y_predict)
for i in range(len(y_predict)):
if y_predict[i]>0.51:
y_predict[i]=int(1)
else:
y_predict[i]=int(0)
y_predict = [int(item) for item in y_predict]
print(y_predict)
print(subdata)
submission = pd.DataFrame({
"id":Id,
"predictrion": y_predict
})
submission.to_csv("one1.csv",index=None,header=None)
来源：网络，如侵，删
