逻辑回归临床预测模型lasso回归变量筛选roc曲线定制Delong检验

from os import name from time import process_time_ns import pandas as pd import numpy as np import matplotlib.pyplot as plt from sklearn.model_selection import train_test_split from sklearn.linear_model import LassoCV,Lasso,LogisticRegression from sklearn.preprocessing import StandardScaler from sklearn.metrics import roc_curve,auc import seaborn as sns import os def get_result(path): data = pd.read_excel(path) n = len(data)+1 x = data.iloc[:,1:n] x = pd.DataFrame(x) x = StandardScaler().fit_transform(x) y = data.iloc[:,0] alphas = np.logspace(-3,1,50) model_lassoCV = LassoCV(alphas = alphas,cv = 10, max_iter = 100000).fit(x,y) lamda = model_lassoCV.alpha_ coef = pd.Series(model_lassoCV.coef_, index = data.columns[1:n]) coefplot = pd.DataFrame() coefplot = pd.DataFrame(coef) coefplot.columns = ['weight'] coefplot = coefplot[coefplot['weight'] !=0] coefplot['index'] = coefplot.index #plt.style.use('ggplot') print(coefplot) plt.figure() p = sns.barplot(x = 'index' , y = 'weight', data=coefplot ) index = coef[coef != 0].index lrx = data.iloc[:,1:n][list(index)] x_train,x_test,y_train,y_test = train_test_split(lrx, y, test_size=0.3, random_state=0) clf = LogisticRegression(max_iter=5000) y_score = clf.fit(x_train,y_train).predict_proba(x_test)[:,1] y_score_data = pd.DataFrame(y_score,index=y_test.index) y_score_train = clf.fit(x_train,y_train).predict_proba(x_train)[:,1] y_score_train = pd.DataFrame(y_score_train,index=y_train.index) train_res = pd.concat([y_train,y_score_train],axis=1) res = pd.concat([y_test,y_score_data],axis=1) fpr,tpr,threshold = roc_curve(y_test,y_score) roc_auc = auc(fpr,tpr) print(roc_auc) orign = 'result/' + path.split('.')[0] if os.path.exists(orign) : name = orign + '/' + 'pre.csv' name0 = orign + '/' + str(lamda) + '.csv' name1 = orign + '/figure.pdf' name2 = orign + '/coef.csv' else: os.makedirs(orign) name = orign + '/' + 'pre.csv' name0 = orign + '/' + str(lamda) + '.csv' name1 = orign + '/figure.pdf' name2 = orign + '/coef.csv' p.figure.savefig(name1, dpi = 400) res.to_csv(name0) train_res.to_csv(name) coefplot.to_csv(name2) files = os.listdir('data') # filelist = [] # for i in files: # if i.split('.')[-1] == 'xlsx' and i.startswith('3'): # filelist.append(i) # print(len(filelist)) # print(files) for file in files: print(file) get_result(file) #get_result('3d_IVIM_ADC基线.xlsx')