Remove outliers from a numpy array using IQR

filter_outliers_iqr.py

import numpy as np

# `data` is expected to be a numpy array
def filter_outliers_iqr(data, iqr_factor=1.5):
    quartile_1, quartile_3 = np.percentile(data, (25, 75))
    iqr = quartile_3 - quartile_1
    lw_bound = quartile_1 - (iqr * iqr_factor)
    up_bound = quartile_3 + (iqr * iqr_factor)

    return data[data >= lw_bound & data <= up_bound]


# same as above but adapted for datasets with features
# and target values (X, Y).
# `data_x` and `data_y` are expected to be a numpy array
def filter_outliers_iqr(data_x, data_y, iqr_factor=1.5):
    quartile_1, quartile_3 = np.percentile(data_y, (25, 75))
    iqr = quartile_3 - quartile_1
    lw_bound = quartile_1 - (iqr * iqr_factor)
    up_bound = quartile_3 + (iqr * iqr_factor)
    
    mask = (data_y >= lw_bound) & (data_y <= up_bound)
    return data_x[mask], data_y[mask]