我有这个简单的时间序列
In [1]: df = pd.DataFrame({'fire': [1, 1, 1]},
...: index=pd.to_datetime([
...: '2016-03-16 23:20:10',
...: '2016-03-16 23:28:58',
...: '2016-03-16 23:38:15']))
...:
In [2]: df
Out[2]:
fire
2016-03-16 23:20:10 1
2016-03-16 23:28:58 1
2016-03-16 23:41:15 1
我想在1分钟前对其进行降采样，然后添加另一个名为 fire_in_the_next_5_minutes 的列。重采样很容易完成，但是我找不到将回填限制为仅前5行的方法。我得到的最接近的数据是：
In [3]: df = df.resample('1min').mean()
...: df['fire_in_the_next_5_minutes'] = df['fire'].fillna(method='backfill')
...:
In [4]: df
Out[4]:
fire fire_in_the_next_5_minutes
2016-03-16 23:20:00 1.0 1.0
2016-03-16 23:21:00 NaN 1.0 <-- should remain NaN
2016-03-16 23:22:00 NaN 1.0 <-- should remain NaN
2016-03-16 23:23:00 NaN 1.0
2016-03-16 23:24:00 NaN 1.0
2016-03-16 23:25:00 NaN 1.0
2016-03-16 23:26:00 NaN 1.0
2016-03-16 23:27:00 NaN 1.0
2016-03-16 23:28:00 1.0 1.0
2016-03-16 23:29:00 NaN 1.0 <-- should remain NaN
2016-03-16 23:30:00 NaN 1.0 <-- should remain NaN
2016-03-16 23:31:00 NaN 1.0 <-- should remain NaN
2016-03-16 23:32:00 NaN 1.0 <-- should remain NaN
2016-03-16 23:33:00 NaN 1.0
2016-03-16 23:34:00 NaN 1.0
2016-03-16 23:35:00 NaN 1.0
2016-03-16 23:36:00 NaN 1.0
2016-03-16 23:37:00 NaN 1.0
2016-03-16 23:38:00 1.0 1.0
是否可以通过其他方式回填，而不是使用fillna方法？
答案 0 :(得分：2)
将public class MainViewModel extends AndroidViewModel implements RecyclerViewAdapter.OnItemClickListener {
private TaskRepository taskRepository;
private LiveData<List<Task>> allTasks;
public final ObservableField<String> description = new ObservableField<>();
public final ObservableField<String> date = new ObservableField<>();
public final ObservableField<String> time = new ObservableField<>();
private String TAG;
private int ID;
private Task mTask;
public MainViewModel(@NonNull Application application) {
super(application);
taskRepository = new TaskRepository(application);
allTasks = taskRepository.getAllTasks();
}
public LiveData<List<Task>> getAllTasks() {
return allTasks;
}
public void deleteTask(Task task) {
taskRepository.delete(task);
}
public void getDataToEdit() {
editTask(description.get(), date.get(), time.get());
}
public void editTask(String description, String date, String time) {
Task task = new Task(description, date, time);
task.setId(ID);
taskRepository.update(task);
Log.d(TAG, "MVM editTask " + task.getId());
}
@Override
public void onItemClick(View view, Task item) {
Log.d(TAG, "MVM " + item.getDescription() + " / " + item.getDate() + " / " + item.getTime() + " / " + mTask);
description.set(item.getDescription());
date.set(item.getDate());
time.set(item.getTime());
ID = item.getId();
mTask.setId(item.getId());
if (mTask != null) {
mTask.setId(item.getId());
}
}
与bfill一起使用
limit答案 1 :(得分：0)
您需要为limit提供fillna参数：
df['fire_in_the_next_5_minutes'] = df['fire'].fillna(method='backfill', limit=5)