卡方分箱

#卡方分箱代码
def ChiMerge(train_new, feature, chuShiQieXiang, xiangShu):
    d1 = pd.DataFrame({'X':train_new[feature], 'Y':train_new.iloc[:,0], feature:pd.qcut(train_new[feature],chuShiQieXiang)})
    d2 = d1.groupby(feature,as_index=True)
    d3=pd.concat([d2.sum()[['Y']],d2[['Y']].count()-d2.sum()[['Y']]],axis=1)
    d3.columns=['bad','good']
    d3['index']=d3.index
    d3.index = range(len(d3))
    while(len(d3)>xiangShu):
        X2=[]
        for i in range (0,len(d3)-1):
            chi2,p,dof,expect=stats.chi2_contingency(d3.iloc[i:i+2,0:2])
            X2.append(chi2)
        ind=X2.index(min(X2))
        chaR=d3.iloc[ind,0:2]+d3.iloc[ind+1,0:2]
        charu=pd.DataFrame({'bad':chaR['bad'],'good':chaR['good'],'index':[pd.Interval(d3.iloc[ind]['index'].left,d3.iloc[ind+1]['index'].right)]})
        d3.drop(d3.iloc[ind:ind+2,:].index,inplace=True)
        heBing=pd.concat([d3,charu])
        heBing=heBing.sort_values('index')
        heBing.index = range(len(heBing))
        d3=heBing
    d3.columns=['bad','good',feature]
    return d3