I am working on my first Full Machine Learning, I am now trying to process raw data, to transform it to supervised learning.
In one of the steps, in order to resamble the data : I have used the resamble, then the apply method and passed The sampling dictionary specifies aggregation functions for different columns.
as the following :

import pandas as pd
from glob import glob

# --------------------------------------------------------------
# Turn into function
# --------------------------------------------------------------

files = glob("../../data/raw/MetaMotion/MetaMotion/*.csv")
data_path = "../../data/raw/MetaMotion/MetaMotion"

def read_data_from_files(files):
    
    acc_df = pd.DataFrame()
    gyr_df = pd.DataFrame()
    
    acc_set = 1
    gyr_set = 1
    
    for f in files:   
        participant = (f.split("-")[0].replace(data_path, ""))[-1]
        label = f.split("-")[1]
        category = f.split("-")[2].rstrip("123").rstrip("_MetaWear_2019")

        df = pd.read_csv(f)
        
        df["participant"] = participant
        df["label"] = label
        df["category"] = category

        if "Accelerometer" in f:
            df["set"] = acc_set
            acc_set =+ 1
            acc_df = pd.concat([acc_df, df])
            
        if "Gyroscope" in f:
            df["set"] = gyr_set
            gyr_set =+ 1
            gyr_df = pd.concat([gyr_df, df])

        
    acc_df.index = pd.to_datetime(acc_df["epoch (ms)"], unit="ms")
    gyr_df.index = pd.to_datetime(gyr_df["epoch (ms)"], unit="ms")
    
    del acc_df["epoch (ms)"]
    del acc_df["time (01:00)"]
    del acc_df["elapsed (s)"]

    del gyr_df["epoch (ms)"]
    del gyr_df["time (01:00)"]
    del gyr_df["elapsed (s)"] 

    return acc_df, gyr_df


acc_df, gyr_df = read_data_from_files(files)

# --------------------------------------------------------------
# Merging datasets
# --------------------------------------------------------------


data_merged = pd.concat([acc_df.iloc[:,:3], gyr_df], axis=1)
data_merged.dropna()
data_merged.info()


#rename columns
data_merged.columns = {
    "acc_x",
    "acc_y",
    "acc_z",
    "gyr_x",
    "gyr_y",
    "gyr_z",
    "label",
    "category",
    "participant",
    "set",
}


# --------------------------------------------------------------
# Resample data (frequency conversion)
# --------------------------------------------------------------

# Accelerometer:    12.500HZ
# Gyroscope:        25.000Hz
 
sampling = {
    "acc_x": "mean",
    "acc_y": "mean",
    "acc_z": "mean",
    "gyr_x": "mean",
    "gyr_y": "mean",
    "gyr_z": "mean",
    "label": "last",
    "category": "last",
    "participant": "last",
    "set": "last",
}

(data_merged[:1000].resample(rule="200ms")).apply(sampling)
resampled_data = data_merged[:1000].resample(rule="200ms").agg(sampling, numeric_only=False)