Liblinear2KaggleSubmission.py

# -- coding: UTF-8 

"""
2016年2月15日
读取liblinear的输出文件，形成kaggle格式的提交方案

liblinear 在指明输出概率时，格式类似这样：
head liresult.txt 
labels 0 1
0 0.506272 0.493728
0 0.506272 0.493728
0 0.506272 0.493728
0 0.506272 0.493728
0 0.506272 0.493728
0 0.506272 0.493728
0 0.506272 0.493728
1 0.5 0.5
0 0.506272 0.493728

test_fn 是测试文件名，需要提取第一个字段（id）
li_fn 是liblinear预测结果文件名,需要提取第三列的正分类概率
res_fn 是最终输出的符合kaggle规则的结果文件名

"""
from itertools import izip

"""
这种实现用iterator遍历一个test文件，同时手工用next遍历相应的linear结果文件
这个结果当是不正确，因为有多余的空行
"""
def toKaggleSubmission(test_fn, li_fn, res_fn):
	with open(test_fn,"rb") as tf, open(li_fn,"rb") as lf, open(res_fn,"w") as rf:
		ti = iter(tf)
		li = iter(lf)
		next(li)
		next(ti)
		rf.write("id,click\n") # header 

		for line in ti:
			sid = line.split(",")[0]
			tline = next(li)
			p = tline.split(" ")[2]
			rf.write("{0},{1}\n".format(sid,p))

"""
这个实现使用izip把两个iterator结合到一块，一起遍历，看起来比较的优雅
"""
def toKaggleSubmission(test_fn, li_fn, res_fn):
	with open(test_fn,"rb") as tf, open(li_fn,"rb") as lf, open(res_fn,"w") as rf:
		ti = iter(tf)
		li = iter(lf)
		next(li)
		next(ti)
		rf.write("id,click\n") # header 

		x = 0
		for tline,line in izip(ti,li):
			sid = tline.split(",")[0]
			p = line.split(" ")[2]
			if x<10:
				x+=1
				print "p={0}".format(p)
			rf.write("{0},{1}".format(sid,p))
			# curious about why here we don't need \n in write

"""
这是分别遍历两个文件提取各自需要的字段，再第三次遍历结合成想要的结果写入
"""
def toKaggleSubmission(test_fn, li_fn, res_fn):
	with open(test_fn,"rb") as tf, open(li_fn,"rb") as lf, open(res_fn,"w") as rf:
		rf.write("id,click\n") # header 

		ids = []
		for line in tf:
			sid = line.split(",")[0]
			ids.append(sid)

		ps = []
		for line in lf:
			p = line.split(" ")[2]
			ps.append(p)

		for i in range(1,len(ids)):
			rf.write("{0},{1}".format(ids[i],ps[i]))

toKaggleSubmission("test.csv","liresult.txt", "submi.csv")

"""
我不明白的是为什么写csv头的那句需要换行符，在里面的又不需要换行符。想想会不会是split(",")的结果中最后一个元素是带换行符的？！
这就可以解释为什么中途尝试输出p=xxx的时候也是有换行
string有splitlines([keepends])方法把字符串分成多个行的数组，
“Line breaks are not included in the resulting list unless keepends is given and true.”
而通常split的是包含换行的
"""