中兴捧月-迪杰特斯拉

import numpy as np
import pandas as pd
PATH='./train_step2.csv'
TEST='./test_step2.csv'
RELT = 'result_step2.txt'
# 训练数据读取
df = pd.read_csv(PATH)
# 待测试数据读取
ts = pd.read_csv(TEST)
# 整理数据
df1 = df.pivot(index='id', columns='date', values='value')
# 现将待填写数据填补一列
ts['value']='nan'
# 将需要填写的id及其前标签提取出来
index_ts =  ts[~ts.duplicated('id')]['id']
m = 0
toData = []
for i in range(index_ts.index.size):
    # 提取目的id
    target = index_ts.iloc[i]
    # 拿出最后五个数据的describe
    des = df1.loc[target].tail(5).describe()
    # 为了避免干扰值，最高和最低都不用
    tailValue = (des['50%'] + des['75%'])/2
    # 如果所得值为nan则赋值为nan 
    if(~np.isnan(tailValue)):
        toVal = tailValue
    else:
        toVal = 'nan'
    # 将值填入
    for j in range(91):
        toData.append(toVal)
# 将值填入
ts['value'] = toData
ts.to_csv(RELT, index=False)
#这版分最高排38，就用这版，最后还被刷了(报告才是重点)...