Python實(shí)現(xiàn)邏輯回歸(Logistic Regression in Python)
http://www.powerxing.com/logistic-regression-in-python/
python sklearn包——混淆矩陣呈枉、分類報(bào)告等自動(dòng)生成
ROC和AUC在python中metrics上的實(shí)現(xiàn)
python sklearn畫ROC曲線
ROC和AUC在python中metrics上的實(shí)現(xiàn)
ROC曲線-閾值評價(jià)標(biāo)準(zhǔn)(詳細(xì)解釋)
http://blog.csdn.net/abcjennifer/article/details/7359370
ROC曲線
http://www.cnblogs.com/webRobot/p/6803747.html
R實(shí)現(xiàn)ROC曲線(BEST CUT OFF)
http://blog.sina.com.cn/s/blog_9b332cf401012qht.html
分類器性能評價(jià):圖形方法(2)
https://site.douban.com/182577/widget/notes/10567212/note/348213109/
二分類模型性能評價(jià)(R語言趁尼,logistic回歸,ROC曲線猖辫,lift曲線酥泞,lorenz曲線)
http://chen.yi.bo.blog.163.com/blog/static/150621109201042641952619/
https://wenku.baidu.com/view/f953978302d276a200292e67.html
計(jì)算Youden指數(shù),并選擇其最大的切點(diǎn)為臨界點(diǎn)
y = combos.aim
pred = combos.predict
from sklearn.metrics import roc_curve
fpr, tpr, thresholds = roc_curve(y, pred)
from sklearn.metrics import auc
auc(fpr, tpr)
plt.plot(fpr, tpr)
#最佳閾值 約登指數(shù)
RightIndex=(tpr+(1-fpr)-1)
tpr[1]
index=np.where(RightIndex==max(RightIndex))
tpr_val=tpr[index]
fpr_val=fpr[index]
thresholds_val=thresholds[index] #0.468