Softmax_Sklearn.py

#-*- coding：utf-8 -*-
# &Author AnFany

importsklearnassk
fromIris_DataimportDataassmdata
importnumpyasnp

fromsklearn.linear_modelimportLogisticRegression

sklr=LogisticRegression(multi_class='multinomial', solver='sag', C=200, max_iter=10000)

#格式化输出混淆矩阵
fromprettytableimportPrettyTable
defconfusion(realy, outy, method='Sklearn'):
mix=PrettyTable()
type=sorted(list(set(realy.T[0])), reverse=True)
mix.field_names= [method] + ['预测:%d类'%siforsiintype]
# 字典形式存储混淆矩阵数据
cmdict= {}
forjkjintype:
cmdict[jkj] = []
forhhintype:
hu=len(['0'forjjinrange(len(realy)) ifrealy[jj][0] ==jkjandouty[jj][0] ==hh])
cmdict[jkj].append(hu)
# 输出表格
forfuintype:
mix.add_row(['真实:%d类'%fu] +cmdict[fu])
returnmix


# 将独热编码的类别变为标识为1，2，3的类别
deftransign(eydata):
ysign= []
forhhineydata:
ysign.append([list(hh).index(1) +1])
returnnp.array(ysign)

# 主函数
if__name__=='__main__':
regre=sklr.fit(smdata[0], transign(smdata[1]).T[0])

predata=np.array([sklr.predict(smdata[0])]).T

print('系数为：\n', np.hstack((sklr.coef_, np.array([sklr.intercept_]).T)).T)

print('混淆矩阵：\n', confusion(transign(smdata[1]), predata))