python numpy 把数据向量化

标签：

某公司的数据处理。

比如有数据：

0 1 20100304 51.14 5 62.2 0 0.0 1 6.82 4.08666666667 4 3
0 0 20100307 2.47333333333 3 75.9333333333 0 0.0 114 13.5666666667 2.86666666667 0 0
0 2 20100318 2.49333333333 3 58.6666666667 0 0.0 1 5.22666666667 1.61333333333 0 0
0 2 20100329 9.14666666667 14 70.6428571429 0 0.0 1 31.84 7.51333333333 11 7
0 0 20100401 3.36666666667 5 71.1 1 75.0 1 1.86 1.29333333333 0 0
0 1 20100401 7.44666666667 11 58.1909090909 0 0.0 51 40.8066666667 12.5066666667 2 1
0 0 20100401 1.0 1 138.0 0 0.0 1 1.0 1.0 0 0
1 0 20100428 1.0 1 85.0 0 0.0 1 0.0 0.0 0 0
1 0 20100429 1.0 1 128.0 0 0.0 1 24.54 3.59333333333 0 0
0 0 20100506 1.92 2 67.0 0 0.0 44 6.65333333333 1.91333333333 4 1

要求：

3、对剩下的数字列，用最大值和最小值的区间，划分十份，每一份就是一个特征，形式：[1,5)、[5,9)...[21,25]，注意最后一个区间闭合了。

4、用上面获取的特征，把原始数据向量化，形式：

1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 

开写：

先下载安装pyExcelerator：http://sourceforge.net/projects/pyexcelerator/，解压后setup.py install

# -*- coding: cp936 -*-

import numpy as np
from numpy import * 
import pprint 
from pyExcelerator import *


#all_col_feature存放特征们
all_col_feature = []

for i in range(2,13):
    #获取[日期]列特征:201003,201004...201408...
    if i == 2 :
        FILENAME = "lost5"
        date = [e[0:6] for e in loadtxt(FILENAME , delimiter = " " , usecols=(2,) , dtype=str)]
        date = list(set(date)) #去重
        date.sort()
        all_col_feature.append(date)
    #获取[数字]列特征：取最大值，最小值，划分10份，每一份就是一个特征,[1,5),[5,9)..[21,25]
    else:
        digit_col = filter(lambda e:e!=0,loadtxt(FILENAME , delimiter = " " , usecols=(i,) , dtype=float))#获取数据后去0
        digit_col.sort()
        max_num = digit_col[-1]
        min_num = digit_col[0]
        block_size = (max_num - min_num) /10
        col_feature = []
        for i in range(11):
            col_feature.append(min_num+i*block_size)
        all_col_feature.append(col_feature)



w = Workbook() #创建一个工作簿    
ws = w.add_sheet('Zero_One_Data') #创建一个工作表
row_index = 0
col_index = 0
for i in range(len(all_col_feature)):
    if i == 0:
        for j in all_col_feature[i]:
            ws.write(row_index,col_index,str(j))
            col_index += 1
    else:
        for j in range(len(all_col_feature[i])-1):
            ws.write(row_index,col_index,str(all_col_feature[i][j])+"~"+str(all_col_feature[i][j+1]))
            col_index += 1
row_index += 1

out = file('out.txt','w')


res = []
for row in loadtxt(FILENAME , delimiter = " " , usecols=(2,3,4,5,6,7,8,9,10,11,12) , dtype=str):
    #数据整理成这个形式：every_row_feature_index = [0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1]
    #比如：第一个0表示在[日期]列的特征下标为0，-1表示NONE值
    #比如：第一个1表示在第二列的特征下标为1，心好累，我这都要注释？
    every_row_feature_index = []
    a = [i for i in range(len(all_col_feature[0])) if row[0][0:6] <= all_col_feature[0][i]]
    every_row_feature_index.append(a[0])
    
    for col in range(1,11):
        b = [i for i in range(len(all_col_feature[col])) if float(row[col]) < float(all_col_feature[col][i])]
        if len(b) == 0:#max_num
            every_row_feature_index.append(len(all_col_feature)-2)
        else:
            every_row_feature_index.append(b[0]-1) #-1表NONE
        
    
    #生成向量
    #就是把every_row_feature_index = [0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1]转成下面的形式：
    #[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
    zero_one_every_row_res = []
    col_index = 0
    for i in range(len(every_row_feature_index)):
        if i == 0: #处理[日期]列
            for j in range(every_row_feature_index[i]):
                out.write('0'+' ')
                ws.write(row_index,col_index,'0')
                col_index += 1
            out.write('1'+' ')
            ws.write(row_index,col_index,'1')
            col_index += 1
            for j in range(every_row_feature_index[i]+1,len(all_col_feature[i])):
                out.write('0'+' ')
                ws.write(row_index,col_index,'0')
                col_index += 1
        else:#处理其他列
            if every_row_feature_index[i] == -1: #-1表NONE值 向量全为0
                for j in range(len(all_col_feature[i])-1):
                    out.write('0'+' ')
                    ws.write(row_index,col_index,'0')
                    col_index += 1
            else:
                for j in range(every_row_feature_index[i]):
                    out.write('0'+' ')
                    ws.write(row_index,col_index,'0')
                    col_index += 1
                out.write('1'+' ')
                ws.write(row_index,col_index,'1')
                col_index += 1
                for j in range(every_row_feature_index[i]+1,len(all_col_feature[i])-1):
                    out.write('0'+' ')
                    ws.write(row_index,col_index,'0')
                    col_index += 1
    out.write('\n')
    row_index += 1
    if row_index == 100:#为了更快看到结果，只设置100行，去掉可得到全部
        break

out.close()
w.save('out.xls')

看看结果：

python numpy 把数据向量化

标签：

原文地址：http://blog.csdn.net/emaste_r/article/details/47836823