🧼 Task 1: Data Cleaning & Preprocessing - Titanic Dataset

This repository contains my submission for Task 1 of the AI & ML Internship: focused on data cleaning and preprocessing using the Titanic dataset.

---

📌 Objective

To clean and prepare raw Titanic dataset for Machine Learning models. This includes:

• Handling missing data
• Encoding categorical features
• Scaling numerical values
• Outlier detection and removal

---

🛠️ Tools & Technologies Used

• Python 🐍
• Pandas & NumPy 📊
• Matplotlib & Seaborn 📈
• Scikit-learn 🔧

---

📁 Dataset

• Titanic Dataset on Kaggle
• File used: titanic.csv

---

🔍 Steps Performed

1. Imported and Explored Data

• Checked data types, null values, and structure of the dataset.

2. Handled Missing Values

• Filled Age with median.
• Filled Embarked with mode.
• Dropped Cabin due to excessive missing values.

3. Categorical Encoding

• Encoded Sex using binary mapping (male=0, female=1).
• Applied One-Hot Encoding to Embarked.

4. Feature Scaling

• Standardized Age and Fare using StandardScaler.

5. Outlier Detection and Removal

• Plotted boxplots for Age and Fare.
• Removed outliers from Fare using the IQR method.

📊 Outlier Boxplot

![Boxplot for Age and Fare]

---

🧠 Key Learnings

• Real-world datasets are messy — preprocessing is a critical step before model training.
• Learned various techniques like median/mode imputation, encoding, and standardization.
• Visualized and cleaned outliers for better model input.

---

❓ Sample Interview Questions Answered

1. Types of missing data: MCAR, MAR, MNAR
2. Handling categorical variables: Label and One-Hot Encoding
3. Normalization vs Standardization:
   • Normalization → scales to [0, 1]
   • Standardization → zero mean, unit variance
4. Outlier detection: Boxplot, IQR
5. Importance of preprocessing: Ensures model quality, avoids bias
6. Data imbalance handling: SMOTE, Resampling, Class weights
7. Encoding methods: One-Hot (independent binary columns), Label (ordinal values)
8. Effect on model accuracy: Strong impact — better preprocessing leads to better models

---

✅ Final Output Snapshot

Here’s a sample of the cleaned dataset:

Data Preprocessing Completed PassengerId Survived Pclass
0 1 0 3
2 3 1 3
3 4 1 1
4 5 0 3
5 6 0 3

                                       Name  Sex       Age  SibSp  Parch  \

0 Braund, Mr. Owen Harris 0 -0.565736 1 0
2 Heikkinen, Miss. Laina 1 -0.258337 0 0
3 Futrelle, Mrs. Jacques Heath (Lily May Peel) 1 0.433312 1 0
4 Allen, Mr. William Henry 0 0.433312 0 0
5 Moran, Mr. James 0 -0.104637 0 0

         Ticket      Fare  Embarked_Q  Embarked_S  

0 A/5 21171 -0.502445 False True
2 STON/O2. 3101282 -0.488854 False True
3 113803 0.420730 False True
4 373450 -0.486337 False True
5 330877 -0.478116 True False