41. What are hyperparameters?
Hyperparameters are external configurations of a model set before training (unlike parameters learned during training). Examples: learning rate, number of trees (in Random Forest), max depth, k in KNN.
42. What is grid search vs random search?
Both are hyperparameter tuning methods:
Grid Search: Exhaustively tests all possible combinations from a defined grid.
Random Search: Randomly selects combinations to test, often faster for large parameter spaces.
43. What are the steps to build a machine learning model?
1. Define the problem
2. Collect and clean data
3. Exploratory Data Analysis (EDA)
4. Feature engineering
5. Split into train/test sets
6. Choose a model
7. Train the model
8. Tune hyperparameters
9. Evaluate on test data
10. Deploy and monitor
44. How do you evaluate model performance?
Depends on the problem type:
Classification: Accuracy, Precision, Recall, F1, ROC-AUC
Regression: RMSE, MAE, R²
Also consider confusion matrix and business context.
45. What is NLP?
NLP (Natural Language Processing) is a field of AI that helps machines understand and interpret human language. Applications: Chatbots, sentiment analysis, translation, summarization.
46. What is tokenization, stemming, and lemmatization?
Tokenization: Splitting text into words or sentences.
Stemming: Trimming words to their root form (e.g., running → run).
Lemmatization: Similar, but more accurate – returns dictionary base form (e.g., better → good).
47. What is topic modeling?
An NLP technique to discover abstract topics in a set of texts.
Common methods: LDA (Latent Dirichlet Allocation), NMF
Used in document classification, summarization, content recommendation.
48. What is deep learning vs machine learning?
Machine Learning: Includes algorithms like regression, decision trees, SVM, etc.
Deep Learning: A subset of ML using neural networks with multiple layers (e.g., CNNs, RNNs).
Deep learning requires more data but can model complex patterns.
49. What is a neural network?
It’s a layered structure of nodes (neurons) that mimic the human brain.
Each node applies weights and activation functions to input and passes it forward.
Used in: Image recognition, speech, NLP, etc.
50. Describe a data science project you worked on
Answer should follow this format:
Problem: What was the goal?
Data: Where did it come from?
Tools: Python, Pandas, Scikit-learn, etc.
Approach: EDA → Feature Engineering → Model → Evaluation
Impact: Quantify improvement (e.g., “increased accuracy by 15%”)