Created
February 16, 2021 03:24
-
-
Save ksv-muralidhar/9bd104692870f740565085a354a2e118 to your computer and use it in GitHub Desktop.
outlier detection
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from sklearn.datasets import load_wine | |
import pandas as pd | |
import numpy as np | |
import matplotlib.pyplot as plt | |
import seaborn as sns | |
data = pd.DataFrame(load_wine()["data"],columns=load_wine()["feature_names"]) | |
data.head() |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
data.plot(kind="box",subplots=True,layout=(7,2),figsize=(15,20)); |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#FUNCTION TO IDENTIFY OUTLIERS USING IQR METHOD | |
def iqr_outlier(x,factor): | |
q1 = x.quantile(0.25) | |
q3 = x.quantile(0.75) | |
iqr = q3 - q1 | |
min_ = q1 - factor * iqr | |
max_ = q3 + factor * iqr | |
result_ = pd.Series([0] * len(x)) | |
result_[((x < min_) | (x > max_))] = 1 | |
return result_ | |
#SCATTER PLOTS HIGHLIGHTING OUTLIERS CALCULATED USING IQR METHOD | |
fig, ax = plt.subplots(7, 2, figsize=(20, 30)) | |
row = col = 0 | |
for n,i in enumerate(data.columns): | |
if (n % 2 == 0) & (n > 0): | |
row += 1 | |
col = 0 | |
outliers = iqr_outlier(data[i], 1.5) | |
if sum(outliers) == 0: | |
sns.scatterplot(x = np.arange(len(data[i])), y = data[i], ax = ax[row, col], legend=False, color = 'green') | |
else: | |
sns.scatterplot(x = np.arange(len(data[i])), y = data[i], ax = ax[row, col], hue = outliers, palette = ['green','red']) | |
for x,y in zip(np.arange(len(data[i]))[outliers == 1], data[i][outliers == 1]): | |
ax[row,col].text(x = x, y = y, s = y, fontsize = 8) | |
ax[row,col].set_ylabel("") | |
ax[row,col].set_title(i) | |
ax[row,col].xaxis.set_visible(False) | |
if sum(outliers) > 0: | |
ax[row,col].legend(ncol=2) | |
col += 1 | |
ax[row,col].axis('off') | |
plt.show() |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#FUNCTION TO DETECT OUTLIERS USING Z-SCORE METHOD | |
def zscore_outlier(x,lb,ub): | |
zscore = ((x - x.mean()) / x.std()).copy() | |
result_ = pd.Series([0] * len(x)) | |
result_[((zscore < lb) | (zscore > ub))] = 1 | |
return result_ | |
#PLOTTING A SCATTER PLOT AND HIGHLIGHTING THE OUTLIERS DETECTED BY Z-SCORE METHOD | |
fig, ax = plt.subplots(7, 2, figsize=(20, 30)) | |
row = col = 0 | |
for n,i in enumerate(data.columns): | |
if (n % 2 == 0) & (n > 0): | |
row += 1 | |
col = 0 | |
outliers = zscore_outlier(data[i], -3, 3) | |
if sum(outliers) == 0: | |
sns.scatterplot(x = np.arange(len(data[i])), y = data[i], ax = ax[row, col], legend=False, color = 'green') | |
else: | |
sns.scatterplot(x = np.arange(len(data[i])), y = data[i], ax = ax[row, col], hue = outliers, palette = ['green','red']) | |
for x,y in zip(np.arange(len(data[i]))[outliers == 1], data[i][outliers == 1]): | |
ax[row,col].text(x = x, y = y, s = y, fontsize = 8) | |
ax[row,col].set_ylabel("") | |
ax[row,col].set_title(i) | |
ax[row,col].xaxis.set_visible(False) | |
if sum(outliers) > 0: | |
ax[row,col].legend(ncol=2) | |
col += 1 | |
ax[row,col].axis('off') | |
plt.show() |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
def euclidean_distance_outlier(x,cutoff): | |
result_ = pd.Series([0] * len(x)) | |
data_mean = x.mean() # mean of data | |
dist = np.sqrt(np.sum(((x-data_mean) ** 2),axis=1)) #Euclidean distande | |
dist_mean = dist.mean() #mean of the distances | |
dist_zscore = np.abs((dist - dist_mean) / dist.std())#z-score of the distances | |
result_[((dist_zscore > cutoff))] = 1 | |
return result_ | |
euc_d = data[["malic_acid","magnesium"]].copy() | |
d['outlier'] = euclidean_distance(d,3) | |
sns.scatterplot(x="malic_acid",y="magnesium",data=d,hue="outlier",palette=["green","red"]) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment