yueeeeeeee / bert4nilm Goto Github PK

It's not clear how to obtain params claimed in the paper specifically for UK-DALE, kettle the paper promises MAE = 6.82 but under default settings for UK-DALE I am getting MAE=22 and also much worse F1....

Hi @Yueeeeeeee ,

I have one doubt regarding the minimum on and minimum off duration that you have used in the utils.py. I believe it is the total num of points that is in minimum on and off duration right and it will depend in sampling rate can you please help me with this.
Also i am trying out the same model for dataport dataset with sampling=15min, can you please help me how i decide the lambda parameter for appliance like.. air conditioner and fridge as well as minimum on and off duration..

the bert is pre-train , so it need to fine-tune after ?

Hello.How to plot the energy consumption cueve of test results using Python?

While reproducing the issue, I encountered the following error. It appears that when processing house 1, you are assigning house_data to entire_data. Later, when attempting to append to it, an error occurs. I suspect that concatenation might be a better approach here. Here's an improved explanation:

When processing data for house 1, you initialize entire_data with house_data. Later, when you try to append new data to entire_data, an error occurs because the append method does not modify the original data structure in place as expected. Instead, it's better to use concatenation to merge the data.

entire_data = entire_data.append(house_data, ignore_index=True)
              ^^^^^^^^^^^^^^^^^^

       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

AttributeError: 'DataFrame' object has no attribute 'append'. Did you mean: '_append'?

           
                if house_id == self.house_indicies[0]:
                    entire_data = house_data
                else:
                    entire_data = entire_data.append(
                        house_data, ignore_index=True)

                entire_data = entire_data.dropna().copy()
                entire_data = entire_data[entire_data['aggregate'] > 0]
                entire_data[entire_data < 5] = 0
                entire_data = entire_data.clip(
                    [0] * len(entire_data.columns), self.cutoff, axis=1)

            return entire_data.values[:, 0], entire_data.values[:, 1:]