-
1
Introduction.
-
2
Show your ID card.
-
3
Explain your notebook/project.
-
4
Explain your notebook as a story.
-
5
Explain your project flow from start to finish.
-
6
Explain the machine learning models you used.
-
7
Explain the preprocessing pipeline you used.
-
8
Explain the hyperparameter tuning process.
-
9
What is the cv parameter in GridSearchCV/RandomizedSearchCV?
-
10
How does Cross Validation work?
-
11
What is K-Fold Cross Validation?
-
12
Why is a validation set needed?
-
13
How many iterations/model fits will be performed based on your hyperparameter tuning?
-
14
Difference between GridSearchCV and RandomizedSearchCV.
-
15
How do you identify whether a model is overfitting or underfitting?
-
16
How do you reduce overfitting?
-
17
What happens if the correlation between features is high?
-
18
How many models did you try?
-
19
Did you try different preprocessing techniques?
-
20
Did you try different models?
-
21
What resources did you use?
-
22
What other approaches did you try?
-
23
What difficulties did you face while doing the project?
-
24
Have you used any sklearn Pipelines?
-
25
Explain your missing value handling.
-
26
How did you handle numerical missing values?
-
27
How did you handle class imbalance?
-
28
Why did you use One-Hot Encoding?
-
29
Have you used scaling?
-
30
Why did you use StandardScaler?
-
31
How does StandardScaler work?
-
32
Difference between StandardScaler and other scaling techniques.
-
33
What is scaling?
-
34
Why didn't you use scaling?
-
35
What is the learning rate parameter?
-
36
Which scoring metric did you use to compare models?
-
37
Why did you use F1-Score, Precision, and Recall instead of RMSE?
-
38
What happens if RMSE is used for Logistic Regression?
-
39
Explain the ROC Curve.
-
40
Explain the Confusion Matrix.
-
41
Point out True Positive (TP) and True Negative (TN) in the Confusion Matrix.
-
42
What is log1p?
-
43
What is the mean and standard deviation after StandardScaler?
-
44
Coding: Implement RandomizedSearchCV for Logistic Regression.
-
45
Coding: Train an ElasticNet model and evaluate its performance.
-
46
Coding: Train a Linear Regression model on the raw (non-preprocessed) dataset.
-
47
Coding: Load the training dataset and print the first 6/7/10 rows.
-
48
Coding: Filter rows where ManufactureYear is greater than a given year.
-
49
Coding: Print unique/value counts of VendorPartnerID.
-
50
Coding: Find the proportion of each category (cardinality) in a feature.
-
51
Coding: Load the Iris dataset and train a Ridge Regression model.
-
52
Coding: Load the California Housing dataset.