Quiz: Evaluation and Tuning

5 questionsNo time limitPass: 60%+10 XP on first pass

Why models that ace the training data fail in real life, how to see overfitting, cross-validation for honest scores, scaling and encoding inside a scikit-learn Pipeline (no data leakage), and tuning with GridSearchCV.

Free account · see your score, answers and explanations, earn XP

Question 1
Training accuracy is 100% but test accuracy is 72%. What is happening?
UnderfittingOverfittingData leakage into the test setPerfect generalisation
Question 2
What does 5-fold cross-validation do?
Trains on 5 rows at a timeTrains 5 times, each time testing on a different fifth of the data, and averages the scoresTests the model on 5 new datasetsRepeats the same split 5 times
Question 3
Why put StandardScaler inside a Pipeline?
It runs fasterSo scaling is learned only from the training data in each fold, avoiding leakagePipelines can't contain models otherwiseTo convert text to numbers
Question 4
Which is an example of data leakage when predicting placement?
Using CGPA as a featureUsing "offer_letter_received" as a featureUsing cross-validationScaling inside a pipeline
Question 5
When should you use the test set?
For every tuning experimentBefore splitting the dataOnce, at the end, to report final performanceNever

Skills101

Skills101 — practical coding courses in Hindi and English: DSA, Python, Java Full Stack and more. Structured lessons, real projects, free tutorials and certificates.

Explore

All courses
Free tutorials
Practice quizzes
Mock tests
Contests
Hackathons
Leaderboard
Interview experiences
Student tools
Online compilers
Tech dictionary
Project showcase
Blog
Corporate training

Contact

info@skills101.in
+91 86199 23140
WhatsApp us
Follow the WhatsApp Channel
Join the WhatsApp group
83-B, Ram Vihar Colony, Sanganer, Jaipur, Rajasthan 302029

YouTube
LinkedIn
X
Facebook