特征缺失值用平均值代替

import numpy as np


e14601 = np.loadtxt(open("/home/zlf/Documents/e14601_data_set.csv", "rb"), delimiter=",", skiprows=0)  

a = e14601.shape;
print(a[0], a[1]);

not0_num = [0 for x in range(a[1]-1)];
not0_sum = [0 for x in range(a[1]-1)];
not0_mean = [0 for x in range(a[1]-1)];
nans = [0 for x in range(a[1]-1)];

for j in range(a[1]-1):
    for i in range(a[0]):
        if e14601[i][j]!=0:
            not0_num[j] = not0_num[j] + 1;
            not0_sum[j] = not0_sum[j] + e14601[i][j];

for j in range(a[1]-1):
    if j<len(not0_num):
        if not0_num[j]==0:
            nans[j] = 1;
            e14601 = np.delete(e14601, j, axis=1);
            not0_num = np.delete(not0_num, j, axis=0);
            not0_sum = np.delete(not0_sum, j, axis=0);

        
a = e14601.shape;
print(a[0], a[1]);

not0_mean = np.divide(not0_sum, not0_num);


for j in range(a[1]-1):
    for i in range(a[0]):
        if e14601[i][j]==0:
            e14601[i][j]=not0_mean[j]
 

np.savetxt('/home/zlf/Documents/e14601_data_set_n0.csv', e14601, delimiter=',')