我用的环境是eclipse中搭建的python开发环境,主要实现基于图像的二分类问题,数据集的遍历以及参数的自动调节,可以参考上一篇博客,在自动调参时,唯一的不同就是参数组合,KNN的参数组合可以参考以下代码:
param_grid = [
{
'weights':['uniform'],
'n_neighbors':[i for i in range(1,8)]
},
{
'weights':['distance'],
'n_neighbors':[i for i in range(1,8)],
'p':[i for i in range(1,4)]
}
]
之后的模型训练和评价,也可以参考上一篇博客,只是将调用的模型改为KNN模型:
classifier = KNeighborsClassifier(weights='distance',n_neighbors=1,p=3)
上一个结果图:
分类效果没有SVM的好。
以下是完整代码:
#KNN算法
import datetime
starttime = datetime.datetime.now()
import numpy as np
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix, classification_report
import os
import cv2
X = []
Y = []
for i in range(0, 2):
#遍历文件夹,读取图片
for f in os.listdir("G:/Download_Data_bishe/leaf_data/pepper_bellSmall/%s" % i):
#打开一张图片并灰度化
Images = cv2.imread("G:/Download_Data_bishe/leaf_data/pepper_bellSmall/%s/%s" % (i, f))
image=cv2.resize(Images,(256,256),interpolation=cv2.INTER_CUBIC)
hist = cv2.calcHist([image], [0,1], None, [256,256], [0.0,255.0,0.0,255.0])
X.append(((hist/255).flatten()))
Y.append(i)
X = np.array(X)
Y = np.array(Y)
#切分训练集和测试集
#X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.3, random_state=1)
#随机率为100%选取其中的30%作为测试集
from sklearn.model_selection import GridSearchCV
# #普通模型训练与预测
# clf0 = KNeighborsClassifier(weights='distance',n_neighbors=1,p=1).fit(X_train, y_train)
# predictions0 = clf0.predict(X_test)
# print(confusion_matrix(y_test, predictions0))
# print (classification_report(y_test, predictions0))
# endtime = datetime.datetime.now()
# print (endtime - starttime)
# #调参函数
# param_grid = [
# {
# 'weights':['uniform'],
# 'n_neighbors':[i for i in range(1,8)]
# },
# {
# 'weights':['distance'],
# 'n_neighbors':[i for i in range(1,8)],
# 'p':[i for i in range(1,4)]
# }
# ]
# scores = ['precision', 'recall']
# for score in scores:
# print("# Tuning hyper-parameters for %s" % score)
# print()
#
# # 调用 GridSearchCV,将 SVC(), tuned_parameters, cv=5, 还有 scoring 传递进去,
# clf = GridSearchCV(KNeighborsClassifier(), param_grid, cv=10,
# scoring='%s_macro' % score) #cv为迭代次数。#基于交叉验证的网格搜索,cv:确定交叉验证拆分策略。
# # 用训练集训练这个学习器 clf
# clf.fit(X_train, y_train)
#
# print("Best parameters set found on development set:")
# print()
#
# # 再调用 clf.best_params_ 就能直接得到最好的参数搭配结果
# print(clf.best_params_)
#
# print()
# print("Grid scores on development set:")
# print()
# means = clf.cv_results_['mean_test_score']
# stds = clf.cv_results_['std_test_score']
#
# # 看一下具体的参数间不同数值的组合后得到的分数是多少
# for mean, std, params in zip(means, stds, clf.cv_results_['params']):
# print("%0.3f (+/-%0.03f) for %r"
# % (mean, std * 2, params))
##交叉验证+roc曲线
import matplotlib.pyplot as plt
from sklearn import svm
from sklearn.metrics import roc_curve, auc
from sklearn.model_selection import StratifiedKFold
#交叉验证
cv = StratifiedKFold(n_splits=10) #导入该模型,后面将数据划分6份
classifier = KNeighborsClassifier(weights='distance',n_neighbors=1,p=3)
# 画平均ROC曲线的两个参数
mean_tpr = 0.0 # 用来记录画平均ROC曲线的信息
mean_fpr = np.linspace(0, 1, 100)
cnt = 0
for i, (train, test) in enumerate(cv.split(X,Y)): #利用模型划分数据集和目标变量 为一一对应的下标
cnt +=1
probas_ = classifier.fit(X[train], Y[train]).predict_proba(X[test]) # 训练模型后预测每条样本得到两种结果的概率
fpr, tpr, thresholds = roc_curve(Y[test], probas_[:, 1]) # 该函数得到伪正例、真正例、阈值,这里只使用前两个
mean_tpr += np.interp(mean_fpr, fpr, tpr) # 插值函数 interp(x坐标,每次x增加距离,y坐标) 累计每次循环的总值后面求平均值
mean_tpr[0] = 0.0 # 将第一个真正例=0 以0为起点
roc_auc = auc(fpr, tpr) # 求auc面积
plt.plot(fpr, tpr, lw=1, label='ROC fold {0:.2f} (area = {1:.2f})'.format(i, roc_auc)) # 画出当前分割数据的ROC曲线
plt.plot([0, 1], [0, 1], '--', color=(0.6, 0.6, 0.6), label='Luck') # 画对角线
mean_tpr /= cnt # 求数组的平均值
mean_tpr[-1] = 1.0 # 坐标最后一个点为(1,1) 以1为终点
mean_auc = auc(mean_fpr, mean_tpr)
plt.plot(mean_fpr, mean_tpr, 'k--',label='Mean ROC (area = {0:.2f})'.format(mean_auc), lw=2)
plt.xlim([-0.05, 1.05]) # 设置x、y轴的上下限,设置宽一点,以免和边缘重合,可以更好的观察图像的整体
plt.ylim([-0.05, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate') # 可以使用中文,但需要导入一些库即字体
plt.title('Receiver operating characteristic example')
plt.legend(loc="lower right")
plt.show()