Python for Data Science Interview Preparation

# CHAPTER 28

Python for Data Science Interview Preparation #

1. Chapter Introduction #

2. Python Core Questions #

1. What is the difference between a List and a Tuple? *Answer:* Both are ordered collections, but Lists [] are mutable (can be changed after creation), while Tuples () are immutable (cannot be changed). Tuples are used for fixed data and use slightly less memory.

2. What is a Dictionary and how does it work? *Answer:* A Dictionary {} is an unordered collection that stores data in Key-Value pairs. It provides O(1) lookup speed for finding a value if you know the exact Key.

3. What is a List Comprehension? (Write an example) *Answer:* It is a compact, "Pythonic" way to create a list from an existing sequence in a single line without writing a standard for loop. *Example:* squared = [x2 for x in range(5)]

4. How do you handle missing data (NaNs) in a Pandas DataFrame? *Answer:* First, I identify them using df.isna().sum(). If the dataset is massive and the missing rows are few, I might drop them using df.dropna(). If I cannot afford to lose data, I impute (fill) them using df.fillna(df.mean()) for numerical data, or the mode for categorical data.

5. What is the difference between loc and iloc? *Answer:* .loc is label-based (selects rows/columns by their explicit index name). .iloc is integer-position based (selects rows/columns by their absolute numerical position, like 0, 1, 2).

6. Write a Pandas command to find the total revenue grouped by Region. *Answer:* df.groupby('Region')['Revenue'].sum()

7. How do you merge two DataFrames? *Answer:* Using pd.merge(df1, df2, on='ID'), which acts exactly like a SQL JOIN, combining the tables based on a shared key.

8. Explain the difference between Supervised and Unsupervised Learning. *Answer:* Supervised learning trains a model on data that has known "answers" or labels (e.g., Regression to predict prices, Classification to predict spam). Unsupervised learning works on data without labels, requiring the algorithm to find hidden structures itself (e.g., Clustering for customer segmentation).

9. What is Data Leakage? *Answer:* Data Leakage occurs when information from outside the training dataset is used to create the model, resulting in overly optimistic accuracy scores. A common cause is applying a StandardScaler to the *entire* dataset before performing the train_test_split.

10. Why must you scale data before using K-Nearest Neighbors (KNN)? *Answer:* KNN relies on calculating the physical distance between data points. If 'Income' is in the tens of thousands and 'Age' is less than 100, the Income feature will completely mathematically dominate the distance calculation. Scaling puts all features on the same numerical range.

11. Why is Accuracy a poor metric for highly imbalanced datasets? *Answer:* If 99% of transactions are legitimate and 1% are fraud, a useless model that just hardcodes "Legitimate" for every transaction will score 99% Accuracy, while failing its only job. You must use Precision, Recall, and the Confusion Matrix to evaluate it properly.

The Challenge: You are given a string containing a messy date: date_str = " 2023/10/15 ". Write a Python one-liner (or short script) to clean it and extract just the year as an Integer.

The Solution:**

date_str = "  2023/10/15  "

# 1. Strip whitespace
# 2. Split by '/' to create a list: ['2023', '10', '15']
# 3. Access the first element [0]
# 4. Cast to integer
year = int(date_str.strip().split('/')[0])

print(year) # 2023

6. Common Mistakes During Interviews #

• Jumping straight to code: When given a data problem, do not immediately write Python. Talk through your process first: "First I would inspect the shape, then look for NaNs, then use groupby..."

• Memorizing functions without understanding concepts: Interviewers care less if you forget the exact parameter name in Scikit-Learn; they care deeply if you don't understand *why* a Train-Test split is mathematically necessary.

7. MCQs #

8. Interview Questions #

• Q: Practice answering this aloud: "Walk me through the lifecycle of a Data Science project, from receiving raw data to deploying a model."

• Q: Practice answering this aloud: "What is a Confusion Matrix, and how do Precision and Recall relate to it?"

9. Summary #

10. Next Chapter Recommendation #