Site icon Codanics

Outliers: Data Science Mein Ghair Mutawaqa Mehman

Outlier Detection and Deadling with them

Outlier Detection and Deadling with them

Data ki duniya mein, outliers woh ghair mutawaqa mehman hain jo aap ki dinner party mein baghair dawat ke aa jate hain. Jab aap samajhte hain ke sab kuch sahi tareeqe se set hai, toh woh aakar harmony ko disturb kar dete hain. Lekin yeh outliers akhir hain kya? Inki ahmiyat kya hai? Aur sab se zaroori, hum inhe kaise handle karte hain? Chaliye, shuru karte hain.

Outliers Kya Hain? 🔍

Outliers are data points that deviate significantly from the rest of the observations in a dataset. Imagine plotting the ages of a group of high school students, and among those teenagers, you find an age of 85. That 85 is an outlier—it doesn’t fit the general trend or expectation.

Outliers Kya Hain? Ek Insightful Kahani 😯

Lahore ki ek masroof galli mein, Ahmed apne doston ke saath chai pee raha tha. ☕

Ali: “Ahmed bhai, kal maine data analyze karte hue ek ajeeb si cheez dekhi. Kuch values bohat zyada alag thi. Kya aapko pata hai woh kya thi?”

Ahmed: “Ali bhai, aap baat kar rahe hain outliers ki. Outliers woh values hain jo baaki data se zyada alag hoti hain. Matlab woh aam data points se hat kar hoti hain.” 😌

Ali: “Outliers? Matlab woh values jo normally expected range se bahar hain?”

Ahmed: “Bilkul! Kabhi kabhi yeh outliers naturally aa jati hain, aur kabhi kuch ghalat input ya measurement ki wajah se. Inko identify karna aur deal karna bohat zaroori hai, kyun ke yeh hamare analysis aur models ko affect kar sakte hain.”

Usman (jo chai ki dukaan ka malik tha): “Toh agar outliers hote hain toh humein unko kya karna chahiye?” 🤔

Ahmed: “Usman bhai, yeh depend karta hai situation par. Kabhi kabhi outliers ko remove karna behtar hota hai, aur kabhi hum unko replace karte hain median ya mean se.”

Ali: “Lekin Ahmed bhai, kaise pata chalega ke koi value outlier hai ya nahi?”

Ahmed: “Ali, bohat se tareeque hain. Jaise ke visual techniques mein scatter plots, box plots waghera. Statistical methods mein z-score ya IQR (Interquartile Range) ki madad se bhi hum outliers identify kar sakte hain.”

Usman: “Acha, toh agar main apni chai ki bikri ki record dekhon aur kuch dinon mein bikri bohat zyada ya bohat kam ho, toh woh bhi outliers ho sakte hain?”

Ahmed: “Bilkul, Usman bhai. Lekin har alag value ko outlier nahi keh sakte. Analyze karna padta hai.”

Gupshup karte karte, Ahmed ne outliers ke baare mein apne doston ko achi tarah se samjhaya. Aur chai ki mazedaar mehfil mein ek aur informative discussion khatam hui. 😄📊

Waise toh outliers humare data mein choti si values hoti hain, lekin inka impact bohat bara hota hai. Isliye, agar aap bhi data science mein naye hain, toh outliers ko samajhna aur unka sahi tareeque se deal karna seekhein. Kyunki data mein choti si baat bhi bari ahemiyat rakhti hai! 🌟📈

Outliers Ko Kaise Pehchaanen?

  1. Visual Tareeqa: Tools jaise scatter plots, box plots, aur histograms behtareen visual madadgar hote hain. Maslan, box plot mein, jo data points whiskers ke bahar hote hain, woh outliers maane ja sakte hain, but we need to know about IQR.
  2. Statistical Tareeqe: Z-score aur IQR (Interquartile Range) method do maqbool statistical tareeqe hain. Z-score batata hai ke ek data point mean se kitne standard deviations door hai. Aam taur par, agar Z-score > 3 ya < -3 ho toh woh outlier maana jata hai.

Inhe Kaise Hatayen? 🛠

  1. Truncation ya Capping: High outliers ke liye, kisi threshold se upar ki value ko maximum cap par set kiya ja sakta hai. Isi tarah, low outliers ke liye, kisi threshold se neeche ki values ko minimum cap par set kiya ja sakta hai.
  2. Transformation: Kabhi-kabhi, logarithms jaise mathematical transformations se outliers ko control kiya ja sakta hai.
  3. Imputation: Outlier ko mean, median, ya mode jaise central tendency measures se replace karein.
  4. Deletion: Agar outlier data entry errors ki wajah se hai ya clear hai ke woh value add nahi karega, toh behtar hai ke aap usey remove kar dein.

Python code to remove outliers

import seaborn as sns
import pandas as pd

# Load the Titanic dataset
titanic = sns.load_dataset('titanic')

# Display the first few rows of the dataset
print(titanic.head())

# Calculate the IQR for the 'age' column
Q1 = titanic['age'].quantile(0.25)
Q3 = titanic['age'].quantile(0.75)
IQR = Q3 - Q1

# Define bounds for the outliers
lower_bound = Q1 - 1.5 * IQR
upper_bound = Q3 + 1.5 * IQR

# Remove outliers
titanic_no_outliers = titanic[(titanic['age'] >= lower_bound) & (titanic['age'] <= upper_bound)]

# Display the first few rows of the dataset without outliers
print(titanic_no_outliers.head())

⚠️ Outliers Ko Ignore Karne Ka Asar

  1. Tircha Analysis: Outliers descriptive aur inferential statistics dono ko skew kar sakte hain, jisse data ka distorted view milta hai.
  2. Machine Learning Models Par Asar: Algorithms, khaas kar linear models, outliers se sensitive hote hain. Woh coefficient estimates aur predictions par drastic asar daal sakte hain.
  3. Gumrahi Paida Karne Wale Assumptions: Data assumptions, jaise normality, outliers ki presence ki wajah se violated ho sakte hain, jisse galat nataij milte hain.

Akhri Baat:

Outliers, halanki woh pareshan kun hote hain, lekin data analysis ka ek ahem hissa hain. Woh anomalies, unique events, ya data collection mein errors ke baare mein zaroori maalumat laate hain. Outliers ko sahi tareeqe se handle karne se mazboot aur reliable analysis ki guarantee hoti hai. Toh, agle dafa jab aap apne dataset mein in uninvited guests ko dekhein, toh aapko bilkul pata hoga ke kya karna hai! 🌟

Read another blog for Data Wrangling Tips and also on Missing Values k Rolay.

Exit mobile version