6 Outliers and Missing Values

6.1 Outliers

Data Mein Ghair Mamooli Cheezon ki Talash 🕵️‍♂️📊

Jab hum data ko samajhne aur us se insights nikalne ki baat karte hain, to kuch aise elements hote hain jo baqi data se hat ke hote hain. Inhe hum ‘anomalies’ ya ‘outliers’ kehte hain. Is chapter mein, hum inhi anomalies ko kaise pehchanein, unka kya asar hota hai, aur unhen kaise handle karein, is par baat karenge.

Outliers woh data points hote hain jo baqi data set se kafi alag hote hain.

Misal: Aapke shehar mein, agar aksar temperature 20°C se 35°C ke darmiyan hota hai, to ek din ka temperature 50°C hona ek outlier hoga.
Ahmiyat: Outliers ko identify karna zaroori hai kyun ke ye kabhi-kabhi data collection mein error ya kisi unusual event ki nishani ho sakte hain.

Outliers are also known as: 1. Abberrant observations 2. Deviants 3. Outlying cases 4. Anomalous points 5. Abnormalities

6.1.1 Types of Outliers

Outliers nine types mein classify kiya ja sakta hai:

Univariate: Ye woh outliers hote hain jo sirf ek variable mein hote hain. For example, agar aapke data mein sirf age variable hai, to age ke outliers univariate outliers honge.
Multivariate: Ye woh outliers hote hain jo ek se zyada variables mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.
Global: Ye woh outliers hote hain jo poore data set mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.
Local: Ye woh outliers hote hain jo sirf ek cluster mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.
Point: Ye woh outliers hote hain jo sirf ek point mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.
Contextual: Ye woh outliers hote hain jo sirf ek cluster mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.
Collective: Ye woh outliers hote hain jo sirf ek cluster mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.
Recurrent: Ye woh outliers hote hain jo sirf ek cluster mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.
Periodic: Ye woh outliers hote hain jo sirf ek cluster mein hote hain. For example, agar aapke data mein age aur income dono variables hain, to age aur income ke outliers multivariate outliers honge.

6.1.2 Causes of Outliers

Outliers ki wajah kuch bhi ho sakti hai. Kuch common causes neeche diye gaye hain:

Data Entry Errors: Data ko enter karte waqt, kisi human error ki wajah se outliers ho sakte hain.
- Misal ke taur par, agar aapke data mein age variable hai, aur kisi ne age ko 100 saal ki jagah 1000 saal enter kar diya, to ye ek outlier hoga.
Measurement Errors: Data ko measure karte waqt, kisi human error ki wajah se outliers ho sakte hain.
- For example, agar aapke data mein height variable hai, aur kisi ne height ko 5 feet ki jagah 50 feet measure kar diya, to ye ek outlier hoga.
Experimental Errors: Data ko experiment karte waqt, kisi human error ki wajah se outliers ho sakte hain.
- For example, agar aapke data mein weight variable hai, aur kisi ne weight ko 50 kg ki jagah 500 kg measure kar diya, to ye ek outlier hoga.
Intentional Outliers: Kisi ne intentionally data mein outliers add kiye hon.
- For example, agar aapke data mein age variable hai, aur kisi ne age ko 100 saal ki jagah 1000 saal enter kar diya, to ye ek outlier hoga.
Data Processing Errors: Data ko process karte waqt, kisi human error ki wajah se outliers ho sakte hain.
- For example, agar aapke data mein age variable hai, aur kisi ne age ko 100 saal ki jagah 1000 saal enter kar diya, to ye ek outlier hoga.
Sampling Errors: Data ko sample karte waqt, kisi human error ki wajah se outliers ho sakte hain.
- For example, agar aapke data mein age variable hai, aur kisi ne age ko 100 saal ki jagah 1000 saal enter kar diya, to ye ek outlier hoga.
Natural Outliers: Data mein outliers ki wajah natural events ho sakte hain.
- For example, agar aapke data mein age variable hai, aur kisi ne age ko 100 saal ki jagah 1000 saal enter kar diya, to ye ek outlier hoga.

6.1.3 Why should we care about Outliers?

Hidden Clues: Outliers humein hidden clues dete hain. Inhe identify kar ke hum kisi hidden pattern ko discover kar sakte hain.
Data Quality: Outliers ki wajah se data quality kam ho jati hai. Inhe identify kar ke hum data quality ko improve kar sakte hain.
Impact Analysis: Outliers ki wajah se humari analysis mein error aa jata hai. Inhe identify kar ke hum analysis ko improve kar sakte hain.
Better Decisions: Outliers ki wajah se humari decisions par bhi asar padta hai. Inhe identify kar ke hum better decisions le sakte hain.
Better Models: Outliers ki wajah se humari models ki accuracy kam ho jati hai. Inhe identify kar ke hum better models bana sakte hain.
Better Insights: Outliers ki wajah se humari insights par bhi asar padta hai. Inhe identify kar ke hum better insights nikal sakte hain.
Better Visualization: Outliers ki wajah se humari visualizations ki quality kam ho jati hai. Inhe identify kar ke hum better visualizations bana sakte hain.
Better Storytelling: Outliers ki wajah se humari storytelling par bhi asar padta hai. Inhe identify kar ke hum better stories bana sakte hain.
Better Data Products: Outliers ki wajah se humari data products ki quality kam ho jati hai. Inhe identify kar ke hum better data products bana sakte hain.
Better Data Science: Outliers ki wajah se humari data science ki quality kam ho jati hai. Inhe identify kar ke hum better data science kar sakte hain.

6.1.4 Detect and remove Outliers

Outliers ko identify karne ke liye, hum kuch techniques use karte hain. In techniques ko hum ‘Outlier Detection Techniques’ kehte hain. In techniques mein se kuch neeche diye gaye hain:

Z-Score
IQR
DBSCAN
Isolation Forest
Local Outlier Factor
Elliptic Envelope
One-Class SVM
Mahalanobis Distance
Robust Random Cut Forest
Histogram-based Outlier Score
K-Nearest Neighbors
K-Means Clustering
Local Correlation Integral
and many more…

Ham sirf Z-Score, IQR or k-means clustering ko dekhenge.

6.1.5 Z-Score Method

Z-Score method mein, hum ye dekhte hain ke koi data point kitne standard deviations (SD) dur hai mean se.

Z-Score ki formula ye hai: \[Z = \frac{x - \mu}{\sigma}\]

Where:
\(Z\): is the Z-Score
\(x\): is the data point
\(\mu\): is the mean of the data
\(\sigma\): is the standard deviation of the data
\(x - \mu\): is the difference between the data point and the mean
\(\frac{x - \mu}{\sigma}\): is the difference between the data point and the mean in terms of standard deviations

Z-Score ki properties ye hain: 1. Z-Score ka mean 0 aur standard deviation 1 hota hai. 2. Z-Score ki value jitni zyada hogi, utna data point mean se zyada dur hoga. 3. Z-Score ki value jitni kam hogi, utna data point mean ke qareeb hoga. 4. Z-Score ki value 3 se zyada ya -3 se kam hogi, to data point outlier hoga.

Z-Score ki values ko interpret karne ke liye, neeche diye gaye table ko dekhein:

Z-Score	Data Point	Interpretation
-3	3 SDs below the mean	Outlier
-2	2 SDs below the mean	Outlier
-1	1 SD below the mean	Outlier
0	Mean	Not an outlier
1	1 SD above the mean	Not an outlier
2	2 SDs above the mean	Not an outlier
3	3 SDs above the mean	Not an outlier

6.1.5.1 Z-Score Method Example in Python

Z-Score method ko Python mein implement karne ke liye, neeche diye gaye steps follow karein:

6.1.5.1.1 Using `numpy`

Run the code below to see the steps.

# Step 1: Import the required libraries
import pandas as pd
import numpy as np

# Step 2: Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 50]})

# Step 3: Calculate the mean and standard deviation
mean = np.mean(data['Age'])
std = np.std(data['Age'])

# Step 4: Calculate the Z-Score
data['Z-Score'] = (data['Age'] - mean) / std

# Step 5: Print the data
print("----------------------------------------")
print(f"Here is the data with outliers:\n {data}")
print("----------------------------------------")
# Step 6: Print the outliers
print(f"Here are the outliers based on the z-score threshold, 3:\n {data[data['Z-Score'] > 3]}")
print("----------------------------------------")
# Step 7: Remove the outliers
data = data[data['Z-Score'] <= 3]

# Step 8: Print the data without outliers
print(f"Here is the data without outliers:\n {data}")

----------------------------------------
Here is the data with outliers:
     Age   Z-Score
0    20 -0.938954
1    21 -0.806396
2    22 -0.673838
3    23 -0.541280
4    24 -0.408721
5    25 -0.276163
6    26 -0.143605
7    27 -0.011047
8    28  0.121512
9    29  0.254070
10   30  0.386628
11   50  3.037793
----------------------------------------
Here are the outliers based on the z-score threshold, 3:
     Age   Z-Score
11   50  3.037793
----------------------------------------
Here is the data without outliers:
     Age   Z-Score
0    20 -0.938954
1    21 -0.806396
2    22 -0.673838
3    23 -0.541280
4    24 -0.408721
5    25 -0.276163
6    26 -0.143605
7    27 -0.011047
8    28  0.121512
9    29  0.254070
10   30  0.386628

6.1.5.1.2 Using `scipy` library

You can also follow the steps below to implement the Z-Score method in Python, using scipy library:

Run the code below to see the steps.

# Import libraries
import numpy as np
from scipy import stats

# Sample data
data = [2.5, 2.7, 2.8, 3.0, 3.2, 3.4, 3.6, 3.8, 4.0, 110.0]

# Calculate the Z-score for each data point
z_scores = np.abs(stats.zscore(data))

# Set a threshold for identifying outliers
threshold = 2.5 
outliers = np.where(z_scores > threshold)[0]

# print the data
print("----------------------------------------")
print("Data:", data)
print("----------------------------------------")

print("Indices of Outliers:", outliers)
print("Outliers:", [data[i] for i in outliers])

# Remove outliers
data = [data[i] for i in range(len(data)) if i not in outliers]
print("----------------------------------------")
print("Data without outliers:", data)

----------------------------------------
Data: [2.5, 2.7, 2.8, 3.0, 3.2, 3.4, 3.6, 3.8, 4.0, 110.0]
----------------------------------------
Indices of Outliers: [9]
Outliers: [110.0]
----------------------------------------
Data without outliers: [2.5, 2.7, 2.8, 3.0, 3.2, 3.4, 3.6, 3.8, 4.0]

6.1.6 IQR Method

IQR method mein, hum ye dekhte hain ke koi data point kitne IQRs dur hai median se.

IQR ki formula ye hai:

\[IQR = Q_3 - Q_1\]

Where:
\(IQR\): is the Interquartile Range
\(Q_3\): is the third quartile
\(Q_1\): is the first quartile
\(Q_3 - Q_1\): is the difference between the third quartile and the first quartile

IQR ki properties ye hain:

IQR ka median 0 aur standard deviation 1 hota hai.
IQR ki value jitni zyada hogi, utna data point median se zyada dur hoga.
IQR ki value jitni kam hogi, utna data point median ke qareeb hoga.

6.1.6.1 IQR Method Example in Python

IQR method ko Python mein implement karne ke liye, neeche diye gaye steps follow karein:

6.1.6.1.1 Using `numpy`

Run the code below to see the steps.

# Step 1: Import the required libraries
import pandas as pd
import numpy as np

# Step 2: Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 50]})

# Step 3: Calculate the first and third quartile
Q1 = np.percentile(data['Age'], 25, interpolation = 'midpoint')
Q3 = np.percentile(data['Age'], 75, interpolation = 'midpoint')

# Step 4: Calculate the IQR
IQR = Q3 - Q1

# Step 5: Calculate the lower and upper bound
lower_bound = Q1 - (1.5 * IQR)
upper_bound = Q3 + (1.5 * IQR)

# Step 6: Print the data
print("----------------------------------------")
print(f"Here is the data with outliers:\n {data}")
print("----------------------------------------")
# Step 7: Print the outliers
print(f"Here are the outliers based on the IQR threshold:\n {data[(data['Age'] < lower_bound) | (data['Age'] > upper_bound)]}")
print("----------------------------------------")
# Step 8: Remove the outliers
data = data[(data['Age'] >= lower_bound) & (data['Age'] <= upper_bound)]

# Step 9: Print the data without outliers
print(f"Here is the data without outliers:\n {data}")

----------------------------------------
Here is the data with outliers:
     Age
0    20
1    21
2    22
3    23
4    24
5    25
6    26
7    27
8    28
9    29
10   30
11   50
----------------------------------------
Here are the outliers based on the IQR threshold:
     Age
11   50
----------------------------------------
Here is the data without outliers:
     Age
0    20
1    21
2    22
3    23
4    24
5    25
6    26
7    27
8    28
9    29
10   30

6.1.7 Clustering Method (K-Means)

Clustering method mein, hum data points ko clusters mein divide karte hain. This can be done using the K-Means clustering algorithm. Where we specify the number of clusters we want to divide the data into. Then we assign each data point to a cluster. Then we calculate the distance of each data point from the centroid of the cluster it belongs to. Then we remove the data points that are farthest from the centroid of the cluster they belong to.

Use the code below to see the steps.

# Import library
from sklearn.cluster import KMeans

# Sample data
data = [[2, 2], [3, 3], [3, 4], [30, 30], [31, 31], [32, 32]]

# Create a K-means model with two clusters (normal and outlier)
kmeans = KMeans(n_clusters=2, n_init=10)
kmeans.fit(data)

# Predict cluster labels
labels = kmeans.predict(data)

# Identify outliers based on cluster labels
outliers = [data[i] for i, label in enumerate(labels) if label == 1]

# print data
print("Data:", data)
print("Outliers:", outliers)
# Remove outliers
data = [data[i] for i, label in enumerate(labels) if label == 0]
print("Data without outliers:", data)

Data: [[2, 2], [3, 3], [3, 4], [30, 30], [31, 31], [32, 32]]
Outliers: [[2, 2], [3, 3], [3, 4]]
Data without outliers: [[30, 30], [31, 31], [32, 32]]

6.1.8 Handling Outliers

Outliers ko handle karne ke liye, hum kuch techniques use karte hain. In techniques ko hum ‘Outlier Handling Techniques’ kehte hain. In techniques mein se kuch neeche diye gaye hain:

Removing the outlier: This is the most common method where all detected outliers are removed from the dataset.
Transforming and binning values: Outliers can be transformed to bring them within a range. Techniques like log transformation or square root transformation can be used.
Imputation: Outliers can also be replaced with mean, median, or mode values.
Separate treatment: In some use-cases, it’s beneficial to treat outliers separately rather than removing or imputing them.
Robus Statistical Methods: Some of the statistical methods to analyze and model the data are less sensitive to outliers and provide more accurate results in the data.

I have explained some of these techniques in the section above.. Where we remove the outliers using the Z-Score, IQR and K-Means clustering methods. You can also use the other techniques by yourself and practice them.

6.1.9 Conclusion

Outliers in a dataset are observations that deviate dramatically from the rest of the data points. They might arise as a result of data gathering mistakes or abnormalities, or they can be real findings that are just infrequent or extraordinary.
If outliers are not appropriately accounted for, they might produce misleading, inconsistent, and erroneous findings. As a result, identifying and dealing with outliers is critical in order to produce accurate and useful data analysis findings.
Outliers may be detected using a variety of methods, including the percentile approach, IQR method, and z-score method. Outliers can be dealt with in a variety of methods, including removal, transformation, imputation, and so on.

6.2 Missing Values

Missing Values Ko Kaise Handle Kiya Jaye? Aur Inhe Handle Karna Kyun Zaroori Hai?” - Data Science Ki Dunia Mein Iska Role 🤔🛠️

Missing values yaani ghaib data se guzarne wala har data scientist ya researcher ko iski ahmiyat aur isse judi mushkilaat ka andaza ho sakta hai. Data Science ki duniya mein, yeh missing values se guzarne ka tajurba aksar humein milta hai. Agar aap mein se kuch khush naseeb hain jo is masle se guzre nahi, toh woh waqai kismat wale hain! 😄 Lekin un logon ke liye jo is masle ka samna karte hain, unko yeh samajhne mein mushkil nahi hoti ke missing values kitne masail paida kar sakti hain.

6.2.1 Naukri, Missing Values aur Aik Bari Ghalti 😢

Lahore ki ek mashhoor company Codanics Solutions mein Ahmed ek talented data scientist tha. Woh apne projects ko hamesha top priority deta tha aur is wajah se us ki company mein bhi bohat izzat thi. 🌟

Ek roz, Ahmed apne doston ke sath lunch kar raha tha. 🍛

Ali (ek aur data scientist): “Ahmed bhai! Suna hai aap ko naya project mila hai?”

Ahmed: “Ji haan, Ali. Mujhe customers ki buying habits analyze karni hai. Lekin data mein kuch missing values hain, mujhe lagta hai koi masla nahi hoga agar main unhein ignore kar doon.” 😕

Ali: “Bhai, kabhi bhi missing values ko ignore mat karo. Yeh choti si baat model ki performance ko kharab kar sakti hai.”

Lekin Ahmed ne Ali ki baat ko nazar andaaz kiya aur apne tareeque se kaam karna shuru kar diya.

Jab model tayyar hua aur us ko real-world data par test kiya gaya, to us ki predictions bilkul bhi sahi nahi thi. 😲 Company ko is wajah se bohat bada nuqsan hua.

CEO, Mr. Usman, ne Ahmed ko apne office mein bulaya. 🏢

Mr. Usman: “Ahmed, humein bohat zyada nuqsan hua hai is project se. Kya masla hai?”

Ahmed: “Sir, maine socha tha ke kuch missing values se koi masla nahi hoga. Lekin mujhe ab samajh aaya hai ke maine ghalat socha.” 😔

Mr. Usman: “Ahmed, aap jante hain data science mein kitni bhi choti ghalti badi problem create kar sakti hai. Mujhe afsos hai, lekin humein aap ko company se nikalna parega.”

Ahmed ko bohat afsos hua. Us ne realize kiya ke kabhi bhi data ko lightly nahi lena chahiye. Woh ghar wapas laut kar Ali ko call kiya. 📞

Ahmed: “Ali, tum sahi keh rahe the. Mujhe company se nikal diya gaya hai.”

Ali: “Afsos hai sun kar. Lekin Ahmed, har galti se humein kuch na kuch seekhne ko milta hai. Aap ab better tareeque se kaam karenge.”

Ahmed ne apni galti se seekha aur woh ab missing values aur data preprocessing par khaas tawajjo dene laga. Chand mahine baad, Ahmed ne ek aur company mein job shuru ki, aur wahan us ne prove kiya ke woh ek maahir data scientist hai. Lekin, us ek ghalti ka sabak us ne hamesha yaad rakha.

Ab agar ap b ahmad ki trah risk lena chahtay hyn tu missing values ko seekhnay se pehlay ap is blog ko ignore kar den, warna agar ap interested hyn tu yaqeen manen ye blog ap ki Data Science or AI ki journey ko bht kamal karne wala hy, I know ap soch rahay hun gay k aisa kia hy is main, Q fir Pola Payen kareay Start? Han Bholay phir tayyar ho?

I know ye nick names hyn magar isi trah or b bht se nick names hyn missing values k, By the way ap apna nick name likhen gay comments main?

6.2.2 Missing Values k ultay naam

Agar ap b aik desi culture ki paidawar hyn tu ap k bhi bht saray ultay naam gay. hai na? like Achoo, Billa, Bhola, Pola, Saji, kala, chitta, mota, chota, kaddu etc., ye main nahi keh raha ap kahin b nazar dorayen tu aisay naaam htay hyn, or kuch tu bht hi adab se pukaray jatay hyn, jaisa k, Paye Kalay. Ab isi trah missing values k bhi naam hyn kaafi jo agar ap ko na pata hun tu ap preshan hun gay. Chalein phir dekhtay hyn!

Missing values ko mukhtalif namon se pukara jata hai, depend karta hai ke context kya hai aur kis domain ya field mein baat ho rahi hai. Lekin, Data Science aur statistics mein commonly istemal hone wale names hain:

NA (Not Available)
NaN (Not a Number): Khaas taur par programming languages jaise ke Python mein pandas library mein istemal hota hai.
Null: Database management systems jaise SQL mein istemal hone wala term hai.
Undefined
Blank ya Empty
Placeholder Values: Kabhi-kabhi kuch default values set ki jati hain jinhein hum recognize kar sakte hain ke yeh actual data nahi hai. Masalan, kisi age field mein -1 ya 999 set karna.
Sentinel Values: Yeh bhi ek tarah ke placeholder values hoti hain jo specific conditions ko represent karte hain.
Dummy Data: Placeholder ya test purpose ke liye istemal hoti hai.
Missing Data: Aam taur se research papers mein istemal hone wala term.

In tamaam terms mein se kuch specific situations ya tools ke liye hote hain, jabke baaz aam istemal ke liye hote hain. Hamesha zaroori hai ke jab aap data ko analyze ya preprocess kar rahe hoon, toh aap in different types ke missing values ko pehchanein aur unhein sahi tareeqay se handle karein.

6.2.3 How to Identify Missing Values?

Missing values ko identify karne ke liye, hum kuch techniques use karte hain. In techniques ko hum ‘Missing Value Detection Techniques’ kehte hain. In techniques mein se kuch neeche diye gaye hain:

Visual Inspection: Data ko visualize kar ke missing values ko identify kiya jata hai.
Descriptive Statistics: Data ki descriptive statistics ko calculate kar ke missing values ko identify kiya jata hai.
Missingno Library: Missingno library ko use kar ke missing values ko identify kiya jata hai.

6.2.3.1 Visual Inspection

Visual Inspection mein, hum data ko visualize kar ke missing values ko identify karte hain.

Use the code below to see the steps.

# Import libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

# Load titanic dataset
data = sns.load_dataset('titanic')

# Visualize the data
plt.figure(figsize=(8, 5))
sns.heatmap(data.isnull(), cbar=False)
plt.show()

Figure 6.1: Visual Inspection of Missing Values of Titanic Dataset. The light colored lines represent missing values.

6.2.3.2 Descriptive Statistics

Descriptive Statistics mein, hum data ki descriptive statistics ko calculate kar ke missing values ko identify karte hain.

Use the code below to see the steps.

# Import libraries
import pandas as pd
import numpy as np
import seaborn as sns
# load titanic dataset
data = sns.load_dataset('titanic')

# calculate missing values
print("----------------------------------------")
print(f"Missing values in each column:\n{data.isnull().sum().sort_values(ascending=False)}")
print("----------------------------------------")
print(f"Percentage of missing values in each column:\n{round(data.isnull().sum() / len(data) * 100, 2).sort_values(ascending=False)}")

----------------------------------------
Missing values in each column:
deck           688
age            177
embarked         2
embark_town      2
survived         0
pclass           0
sex              0
sibsp            0
parch            0
fare             0
class            0
who              0
adult_male       0
alive            0
alone            0
dtype: int64
----------------------------------------
Percentage of missing values in each column:
deck           77.22
age            19.87
embarked        0.22
embark_town     0.22
survived        0.00
pclass          0.00
sex             0.00
sibsp           0.00
parch           0.00
fare            0.00
class           0.00
who             0.00
adult_male      0.00
alive           0.00
alone           0.00
dtype: float64

6.2.3.3 Missingno Library

Missingno library ko use kar ke bhi hum missing values ko identify kar sakte hain.

Use the code below to see the steps.

Code

# Import libraries
import pandas as pd
import numpy as np
import seaborn as sns

df = sns.load_dataset('titanic')
df

	survived	pclass	sex	age	sibsp	parch	fare	embarked	class	who	adult_male	deck	embark_town	alive	alone
0	0	3	male	22.0	1	0	7.2500	S	Third	man	True	NaN	Southampton	no	False
1	1	1	female	38.0	1	0	71.2833	C	First	woman	False	C	Cherbourg	yes	False
2	1	3	female	26.0	0	0	7.9250	S	Third	woman	False	NaN	Southampton	yes	True
3	1	1	female	35.0	1	0	53.1000	S	First	woman	False	C	Southampton	yes	False
4	0	3	male	35.0	0	0	8.0500	S	Third	man	True	NaN	Southampton	no	True
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
886	0	2	male	27.0	0	0	13.0000	S	Second	man	True	NaN	Southampton	no	True
887	1	1	female	19.0	0	0	30.0000	S	First	woman	False	B	Southampton	yes	True
888	0	3	female	NaN	1	2	23.4500	S	Third	woman	False	NaN	Southampton	no	False
889	1	1	male	26.0	0	0	30.0000	C	First	man	True	C	Cherbourg	yes	True
890	0	3	male	32.0	0	0	7.7500	Q	Third	man	True	NaN	Queenstown	no	True

891 rows × 15 columns

Titanic Dataset

# Import libraries
import pandas as pd
import numpy as np
import missingno as msno
import matplotlib.pyplot as plt
import seaborn as snss

# load titanic dataset
data = sns.load_dataset('titanic')
# Visualize the data
msno.matrix(data,labels=True, fontsize=12, width_ratios=(2, 4), color=(0.2, 0.4, 0.6))
plt.show()

6.2.4 Missing Values Handle Karna Kyun Itna Ahem Hai? 🤷‍♂️

Model Ki Accuracy Par Gehra Asar: 💔Missing values ke honay se machine learning models ki accuracy mein kami aati hai, aur iski performance par bhi bura asar hota hai.
Data Ki Mayari Par Sawal: 📉Missing values data ki mayari ko kamzor banate hain, jisse hamare analysis aur faislay mein bhi ghalat fehmiyan paida ho sakti hain.
Model Training Ka Waqt Barh Jata Hai: ⏱️Kabhi-kabhi, missing values ki wajah se model training ka waqt barh jata hai, jo ke resources aur waqt dono ka zaya hai.

6.2.5 Ruku Zara Sabr Karo

Missing values ka hona kisi bhi dataset mein aam baat hai, lekin jab hum decide karte hain ke kisi column ko remove karna chahiye ya nahi, to iska faisla humein kuch factors par depend karta hai:

Data Ki Quantity: Agar aapke paas bohat zyada data hai aur aik specific column mein missing values ki tadad bohat zyada hai (masalan, 70% ya 80%), toh us column ko remove kar dena behtar ho sakta hai, kyun ke us column se faida uthana mushkil ho sakta hai.

Column Ki Importance: Agar missing values wala column aapke analysis ya model ke liye bohat ahem hai, toh us column ko remove karna acha nahi hoga. Aise mein aap missing values ko impute karne ke tareeqe istemal kar sakte hain.

Nature of Data: Kabhi-kabhi, missing values ka hona bhi kuch indicate karta hai. Masalan, kisi survey mein, agar kisi sawal ka jawab nahi diya gaya, toh yeh indicate kar sakta hai ke participant us sawal se comfortable nahi tha. Aise mein, missing value ko hata dena ya replace karna sahi nahi hoga.

Model Ki Sensitivity: Kuch machine learning models missing values ko handle kar sakte hain, jabke kuch models sensitive hoti hain. Aise mein, agar model missing values ke sensitive hai, toh aapko missing values ko handle karna parega.

Type of Data: Numeric data mein missing values ko mean, median ya mode se replace kiya ja sakta hai. Categorical data mein, missing values ko mode ya kisi specific category se replace kiya ja sakta hai.

Aam taur par, agar aapke column mein 50% se zyada data missing hai, toh us column ko consider karna chahiye ke kya usse remove karna behtar rahega ya nahi. Lekin, yeh hard and fast rule nahi hai. Har dataset unique hota hai aur uski requirements bhi alag hoti hain. Is liye, aapko har dataset ke context mein decide karna hoga ke missing values ko kaise handle kiya jaye.

6.2.6 Missing Values Ko Handle Karne Ke Mufassal Tariqay 🧐

6.2.6.1 Maujooda Data Source Se Phir Se Data Hasil Karna: 🔄Agar aap ke paas woh resource maujood hai jahan se aapne data liya tha, toh aap missing values ko wahan se dobara hasil kar sakte hain.

6.2.6.2 Mean, Median, Ya Mode Se Data Ko Impute Karna: 📊Agar aapke paas numerical data hai, toh usmein missing values ko mean ya median se replace kiya jata hai. Wahi, categorical data ke liye mode ka istemal hota hai.

Use following code to see the steps to fill missing values with mean, median or mode in Python:

1. Mean

# Import libraries
import pandas as pd
import numpy as np

# Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, np.nan, 50]})

# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Calculate the mean
mean = data['Age'].mean()

# Replace the missing values with mean
data['Age'] = data['Age'].fillna(mean)
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10   NaN
11  50.0
----------------------------------------
Here is the data without missing value:
           Age
0   20.000000
1   21.000000
2   22.000000
3   23.000000
4   24.000000
5   25.000000
6   26.000000
7   27.000000
8   28.000000
9   29.000000
10  26.818182
11  50.000000

2. Median

# Import libraries
import pandas as pd
import numpy as np

# Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, np.nan, 50]})
# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Calculate the median
median = data['Age'].median()

# Replace the missing values with median
data['Age'] = data['Age'].fillna(median)
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10   NaN
11  50.0
----------------------------------------
Here is the data without missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10  25.0
11  50.0

Mode

# Import libraries
import pandas as pd
import numpy as np

# Create the data categorical data with mode and missing value
data = pd.DataFrame({'Fruit': ['Apple', 'Banana', 'Apple', 'Banana', 'Apple', 'Banana', 'Apple', 'Banana', 'Apple', 'Banana', np.nan, 'Banana']})

# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Find the mode
mode = data['Fruit'].mode()[0]

# Replace the missing values with mode
data['Fruit'] = data['Fruit'].fillna(mode)
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Fruit
0    Apple
1   Banana
2    Apple
3   Banana
4    Apple
5   Banana
6    Apple
7   Banana
8    Apple
9   Banana
10     NaN
11  Banana
----------------------------------------
Here is the data without missing value:
      Fruit
0    Apple
1   Banana
2    Apple
3   Banana
4    Apple
5   Banana
6    Apple
7   Banana
8    Apple
9   Banana
10  Banana
11  Banana

6.2.6.3 Forward Ya Backward Fill Ka Istemal: 🚶‍♂️🏃‍♂️Kuch data sets mein waqt ya tarikh ka silsila hota hai. Aise data sets mein, aik row ke missing value ko pichli ya agli row ki value se replace kiya jata hai.

Use following code to see the steps to fill missing values with forward or backward fill in Python:

1. Forward Fill

# Import libraries
import pandas as pd
import numpy as np

# Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, np.nan, 50]})

# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Replace the missing values with forward fill
data['Age'] = data['Age'].ffill()
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10   NaN
11  50.0
----------------------------------------
Here is the data without missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10  29.0
11  50.0

2. Backward Fill

# Import libraries
import pandas as pd
import numpy as np

# Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, np.nan, 50]})
# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Replace the missing values with backward fill
data['Age'] = data['Age'].bfill()
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10   NaN
11  50.0
----------------------------------------
Here is the data without missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10  50.0
11  50.0

6.2.6.4 KNN Imputation Ka Istemal: 🧑‍🤝‍🧑Yeh ek advanced technique hai jahan missing value ko uske aas-paas ke data points ke average value se replace kiya jata hai. Aise libraries jaise scikit-learn mein yeh method maujood hai.

Use following code to see the steps to fill missing values with KNN imputation in Python:

# Import libraries
import pandas as pd
import numpy as np
from sklearn.impute import KNNImputer

# Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, np.nan, 50]})
# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Initialize the KNNImputer
imputer = KNNImputer(n_neighbors=2)

# Replace the missing values with KNN imputation
data['Age'] = imputer.fit_transform(data[['Age']])
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10   NaN
11  50.0
----------------------------------------
Here is the data without missing value:
           Age
0   20.000000
1   21.000000
2   22.000000
3   23.000000
4   24.000000
5   25.000000
6   26.000000
7   27.000000
8   28.000000
9   29.000000
10  26.818182
11  50.000000

6.2.6.5 Deep Learning Techniques Ka Istemal: 🧠Deep learning techniques jaise autoencoders bhi missing values ko handle karne mein madadgar sabit ho sakte hain.

Use following code to see the steps to fill missing values with deep learning techniques in Python:

# Import libraries
import pandas as pd
import numpy as np
from sklearn.experimental import enable_iterative_imputer
from sklearn.impute import IterativeImputer

# Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, np.nan, 50]})
# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Initialize the IterativeImputer
imputer = IterativeImputer()

# Replace the missing values with deep learning techniques
data['Age'] = imputer.fit_transform(data[['Age']])
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10   NaN
11  50.0
----------------------------------------
Here is the data without missing value:
           Age
0   20.000000
1   21.000000
2   22.000000
3   23.000000
4   24.000000
5   25.000000
6   26.000000
7   27.000000
8   28.000000
9   29.000000
10  26.818182
11  50.000000

6.2.6.6 Simply Delete Kar Dena: ❌Agar aapke data set mein missing values ki tadad bahut kam hai, toh aap us specific row ya column ko bhi delete kar sakte hain.

Use following code to see the steps to delete missing values in Python:

# Import libraries
import pandas as pd
import numpy as np

# Create the data
data = pd.DataFrame({'Age': [20, 21, 22, 23, 24, 25, 26, 27, 28, 29, np.nan, 50]})
# Print the data with missing value
print("----------------------------------------")
print(f"Here is the data with missing value:\n {data}")

# Delete the rows with missing values
data = data.dropna()
print("----------------------------------------")
# Print the data without missing value
print(f"Here is the data without missing value:\n {data}")

----------------------------------------
Here is the data with missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
10   NaN
11  50.0
----------------------------------------
Here is the data without missing value:
      Age
0   20.0
1   21.0
2   22.0
3   23.0
4   24.0
5   25.0
6   26.0
7   27.0
8   28.0
9   29.0
11  50.0

6.2.6.7 Agar main na handle karun tu?

Bachoo Jee! phir tu hargiz model acha kaam nahi kare ga, yehi nahi abhi or suneay!

Agar hum missing values ko nazar andaaz kar dein toh humein kai masail ka samna karna par sakta hai. Yahan kuch masail hain jo arise ho sakti hain:

📉 Model Accuracy Mein Kami: Machine learning models ki accuracy kam ho sakti hai, kyun ke model ko complete information nahi milti.

📊 Ghalat Analysis: Data analysis mein ghalat nataij nikal sakte hain, jo ke decisions par negative asar dal sakta hai.

😕 Model Confusion: Kuch models missing values handle nahi kar pate, jis se model train nahi ho pata ya phir ghalat predictions karta hai.

🤖 Bias in Model: Missing values ki wajah se model mein bias aane ka khatra barh jata hai.

📚 Data ka Ghalat Interpretation: Missing values ki wajah se humare paas adhoori ya ghalat malumat ho sakti hai, jis ki wajah se hum data ko ghalat tareeqe se interpret kar sakte hain.

💾 Storage Issues: Agar missing values ko replace nahi kiya jaye toh storage mein bhi masail ho sakti hain, kyun ke kuch systems missing values ko store nahi kar pate.

🔀 Data Integration Masail: Different sources se aane wale data mein agar missing values hain toh integration mein masail ho sakti hain.

🚫 Features ka Ghalat Selection: Missing values ki presence mein, kuch aham features ko ignore kiya ja sakta hai jin ka model par asar hona chahiye.

🧪 Ghalat Experimental Results: Science ya research projects mein, missing values ki wajah se ghalat experimental nataij aa sakte hain.

😰 Stress aur Extra Kaam: Data scientists ko extra kaam karna par sakta hai tajziyat mein, kyun ke missing values ko identify aur handle karna parta hai.

Is liye, missing values ko handle karna bohat zaroori hota hai ta ke hum upar diye gaye masail se bach saken. 🛠️🔧🔍

6.2.7 Conclusion

Missing Values - Ek Badi Challenge Lekin Ek Behtareen Mauqa Bhi 🌟Missing values se guzarne ka tajurba har data scientist ke liye ek challenge toh hai hi, lekin isse humein yeh bhi seekhne ko milta hai ke hum kaise data ki mayari ko behtar bana sakte hain. Aakhir mein, behtar quality wale data se hi behtar aur zaheen insights aur models tayyar hoti hain.

6.3 Follow us

Main umeed karta hun k ap ko ye chapter ne bht kuch seekhaya ho ga, or agar sach main seekhaya hy then please do support us by sharing this book with your friends and colleagues. Also, do share your feedback with us, so that we can improve our work in future.

6.1 Outliers

6.1.1 Types of Outliers

6.1.2 Causes of Outliers

6.1.3 Why should we care about Outliers?

6.1.4 Detect and remove Outliers

6.1.5 Z-Score Method

6.1.5.1 Z-Score Method Example in Python

6.1.5.1.1 Using numpy

6.1.5.1.2 Using scipy library

6.1.6 IQR Method

6.1.6.1 IQR Method Example in Python

6.1.6.1.1 Using numpy

6.1.7 Clustering Method (K-Means)

6.1.8 Handling Outliers

6.1.9 Conclusion

6.2 Missing Values

6.2.1 Naukri, Missing Values aur Aik Bari Ghalti 😢

6.2.2 Missing Values k ultay naam

6.2.3 How to Identify Missing Values?

6.2.3.1 Visual Inspection

6.2.3.2 Descriptive Statistics

6.2.3.3 Missingno Library

6.2.4 Missing Values Handle Karna Kyun Itna Ahem Hai? 🤷‍♂️

6.2.5 Ruku Zara Sabr Karo

6.2.6 Missing Values Ko Handle Karne Ke Mufassal Tariqay 🧐

6.2.6.1 Maujooda Data Source Se Phir Se Data Hasil Karna: 🔄Agar aap ke paas woh resource maujood hai jahan se aapne data liya tha, toh aap missing values ko wahan se dobara hasil kar sakte hain.

6.2.6.2 Mean, Median, Ya Mode Se Data Ko Impute Karna: 📊Agar aapke paas numerical data hai, toh usmein missing values ko mean ya median se replace kiya jata hai. Wahi, categorical data ke liye mode ka istemal hota hai.

6.2.6.3 Forward Ya Backward Fill Ka Istemal: 🚶‍♂️🏃‍♂️Kuch data sets mein waqt ya tarikh ka silsila hota hai. Aise data sets mein, aik row ke missing value ko pichli ya agli row ki value se replace kiya jata hai.

6.2.6.4 KNN Imputation Ka Istemal: 🧑‍🤝‍🧑Yeh ek advanced technique hai jahan missing value ko uske aas-paas ke data points ke average value se replace kiya jata hai. Aise libraries jaise scikit-learn mein yeh method maujood hai.

6.2.6.5 Deep Learning Techniques Ka Istemal: 🧠Deep learning techniques jaise autoencoders bhi missing values ko handle karne mein madadgar sabit ho sakte hain.

6.2.6.6 Simply Delete Kar Dena: ❌Agar aapke data set mein missing values ki tadad bahut kam hai, toh aap us specific row ya column ko bhi delete kar sakte hain.

6.2.6.7 Agar main na handle karun tu?

6.2.7 Conclusion

6.3 Follow us

6.1.5.1.1 Using `numpy`

6.1.5.1.2 Using `scipy` library

6.1.6.1.1 Using `numpy`