我有一个如下所示的数据框架：
我想创建一个名为“flag”的列，用于单独监测每个RB的“value”列的变化，并且仅当数值增加时，在变化前的那个月标记为1，但前提是变化前的那一个月的RB必须与变化发生的月份相同。因此，我认为简单的移位操作是无法实现的。
我还想创建一个类似的列，当某个RB的数值增加时，在变化前的那个月（如上列所示）、两个月前和三个月前都标记为1，规则保持不变，只有当变化发生的月份和这三个月的RB都相同的情况下，才对这三个月进行“移位”。
回答：
这应该可以实现您想要的效果：
import pandas as pddata = [ {"rb": 111, "date": "01/01/2020", "value": 5}, {"rb": 111, "date": "01/02/2020", "value": 5}, {"rb": 111, "date": "01/03/2020", "value": 4}, {"rb": 111, "date": "01/04/2020", "value": 6}, {"rb": 111, "date": "01/05/2020", "value": 6}, {"rb": 111, "date": "01/06/2020", "value": 6}, {"rb": 111, "date": "01/07/2020", "value": 6}, {"rb": 111, "date": "01/08/2020", "value": 7}, {"rb": 112, "date": "01/01/2020", "value": 3}, {"rb": 112, "date": "01/02/2020", "value": 3}, {"rb": 112, "date": "01/03/2020", "value": 4}, {"rb": 112, "date": "01/04/2020", "value": 4}, {"rb": 112, "date": "01/05/2020", "value": 5}, {"rb": 112, "date": "01/06/2020", "value": 5}, {"rb": 112, "date": "01/07/2020", "value": 5}, {"rb": 112, "date": "01/08/2020", "value": 5}, {"rb": 111, "date": "01/01/2020", "value": 18}, {"rb": 111, "date": "01/02/2020", "value": 18}, {"rb": 111, "date": "01/03/2020", "value": 17}, {"rb": 111, "date": "01/04/2020", "value": 11}, {"rb": 111, "date": "01/05/2020", "value": 13}, {"rb": 111, "date": "01/06/2020", "value": 13}, {"rb": 111, "date": "01/07/2020", "value": 13}, {"rb": 111, "date": "01/08/2020", "value": 13}, {"rb": 112, "date": "01/01/2020", "value": 14}, {"rb": 112, "date": "01/02/2020", "value": 14}, {"rb": 112, "date": "01/03/2020", "value": 17}, {"rb": 112, "date": "01/04/2020", "value": 17}, {"rb": 112, "date": "01/05/2020", "value": 5}, {"rb": 112, "date": "01/06/2020", "value": 5}, {"rb": 112, "date": "01/07/2020", "value": 5}]df = pd.DataFrame(data)df["flag"] = 0for index in range(len(df) - 1): df.loc[index, "flag"] = int(df.loc[index, "rb"] == df.loc[index + 1, "rb"] and df.loc[index, "value"] < df.loc[index + 1, "value"])df["flag_3m"] = 0for index in range(len(df)): try: df.loc[index, "flag_3m"] = int(df.loc[index, "flag_3m"] != 1 and ((df.loc[index, "value"] < df.loc[index + 1, "value"] and df.loc[index, "rb"] == df.loc[index + 1, "rb"]) or (df.loc[index + 1, "value"] < df.loc[index + 2, "value"] and df.loc[index, "rb"] == df.loc[index + 2, "rb"]) or (df.loc[index + 2, "value"] < df.loc[index + 3, "value"] and df.loc[index, "rb"] == df.loc[index + 3, "rb"]))) except: # 粗糙的方法 ;) passprint(df)PS：也许先按rb进行groupby然后检查数据会更容易，但这种方法也应该可以工作。
