我从csv文件读取数据帧,它类似于以下内容:
LIST-1 LIST-2 LIST-3 ... LIST-N
TIME
2017-06-21 00:17:00 NaN [99.221] [42.357, 102.665]
2017-06-21 00:18:00 NaN [50.89] [42.357, 43.125,...]
2017-06-21 00:19:00 NaN [61.50, 76.1] [70.163, 121.486]
2017-06-21 00:20:00 [70.16] NaN NaN
2017-06-21 00:21:00 NaN [102.665] [57.9, 63.66, 68.7...
每行代表一分钟的数据,list_N列的dtype代表对象。我想做:
ALL_LIST
; ALL_LIST
))合并到一个新列表中; 最后,我想得到一个这样的数据框:
TIME LIST 2017-06-21 00:00:00 [99.221,42.357, 42.357, ...] 2017-06-21 00:30:00 [52.328,42.357, 49.169, ...] 2017-06-21 01:00:00 [61.484,42.357, 76.52, ...] 2017-06-21 01:30:00 [76.523,42.357, 121.486, ...]
答案 0 :(得分:1)
我为我的问题找到了一个解决方案。我会写出来并希望看看它是否能提高性能。
all_tt_list['ALL_LIST'] = all_tt_list.apply(lambda x: ','.join(x.dropna()), axis=1)
all_tt_list['ALL_LIST'] = all_tt_list['ALL_LIST'].astype(str).str.replace('[', '')
all_tt_list['ALL_LIST'] = all_tt_list['ALL_LIST'].astype(str).str.replace(']', '')
all_tt_list['ALL_LIST'] = all_tt_list['ALL_LIST'].astype(str).str.split(',')
WAIT_TIME_INTERVAL = 30*60
rng = pd.date_range(date, periods=(24 * 60 * 60 / WAIT_TIME_INTERVAL) + 1, freq=str(WAIT_TIME_INTERVAL) + 'S',
tz='Asia/Shanghai')
for k in range(len(rng)):
if(k == (len(rng)-1)):
continue
period_start = rng[k]
period_end = rng[k+1]
period_df = all_tt_list[all_tt_list.index > period_start]
period_df = period_df[period_df.index < period_end]
period_tt_list = period_df['ALL_LIST'].tolist()
import itertools
period_merged = list(itertools.chain.from_iterable(period_tt_list))
period_merged_s = pd.DataFrame(period_merged, columns=['TT_NUM']).astype(float).astype(int)