Viva prep · Real questions · Student experiences Enroll in Bootcamp

Search viva questions

mlp_level1_viva7_MLP_Viva MLP Level 1 Aug 5, 2026
  1. 1
    Explain your notebook/project.
  2. 2
    Explain your code and approach.
  3. 3
    Have you used Pipeline?
  4. 4
    Have you used ColumnTransformer?
  5. 5
    Why is Pipeline used?
  6. 6
    Explain SimpleImputer.
  7. 7
    Explain One-Hot Encoding.
  8. 8
    Difference between One-Hot Encoding and Label Encoding.
  9. 9
    Which Cross Validation (CV) technique did you use?
  10. 10
    What is XGBoost?
  11. 11
    How does XGBoost work?
  12. 12
    What is overfitting?
  13. 13
    What is underfitting?
level2_81 MLP Level 1 Aug 5, 2026
  1. 1
    Explain your notebook/project.
  2. 2
    Explain the problem statement.
  3. 3
    What approach did you follow to solve the problem?
  4. 4
    What are the top insights from your EDA?
  5. 5
    Explain every graph used in your EDA.
  6. 6
    Why did you choose those plots?
  7. 7
    How can you make your graphs cleaner?
  8. 8
    What did you understand from the describe() output?
  9. 9
    How did you correlate features with the target variable?
  10. 10
    Why did you drop specific columns?
  11. 11
    How did you identify and remove duplicates?
  12. 12
    How did you handle missing values?
  13. 13
    Why did you choose that imputation technique?
  14. 14
    What preprocessing did you perform?
  15. 15
    Why is preprocessing required?
  16. 16
    What feature engineering did you perform?
  17. 17
    Did you create any feature that improved your score significantly?
  18. 18
    Did you create any feature that improved your score significantly?
  19. 19
    How do lag features help?
  20. 20
    How do rolling features help?
  21. 21
    How did you implement lag and rolling features?
  22. 22
    What is Backward Fill?
  23. 23
    What is Forward Fill?
  24. 24
    What encoding techniques did you use?
  25. 25
    Explain One-Hot Encoding.
  26. 26
    Explain Label Encoding.
  27. 27
    Explain Ordinal Encoding.
  28. 28
    Difference between One-Hot Encoding and Label Encoding.
  29. 29
    When should you use Label Encoding?
  30. 30
    What are the drawbacks of One-Hot Encoding?
  31. 31
    What scaling techniques did you use?
  32. 32
    Why is scaling required?
  33. 33
    What happens if scaling is not applied?
  34. 34
    How does StandardScaler work?
  35. 35
    How does MinMaxScaler work?
  36. 36
    Difference between StandardScaler and MinMaxScaler.
  37. 37
    What is the range after applying StandardScaler?
  38. 38
    What is the range after applying MinMaxScaler?
  39. 39
    After applying StandardScaler, what are the new mean and standard deviation?
  40. 40
    What percentage of values lie between -3σ and +3σ in a normal distribution?
  41. 41
    Explain TF-IDF.
  42. 42
    Why did you choose those models?
  43. 43
    Explain the working of each model.
  44. 44
    Explain Logistic Regression.
  45. 45
    Explain the Sigmoid function.
  46. 46
    Write the Sigmoid function formula.
  47. 47
    How does Logistic Regression learn weights?
  48. 48
    What is the loss function of Logistic Regression?
  49. 49
    What changes are required for multiclass Logistic Regression?
  50. 50
    Explain SVM.
  51. 51
    SVM vs Logistic Regression for outliers.
  52. 52
    Explain Decision Tree.
  53. 53
    Explain Random Forest.
  54. 54
    Explain Naive Bayes.
  55. 55
    What are the limitations of Naive Bayes?
  56. 56
    Explain XGBoost.
  57. 57
    Explain LightGBM.
  58. 58
    Difference between XGBoost and LightGBM.
  59. 59
    Why is LightGBM better than other boosting algorithms?
  60. 60
    Explain level-wise vs leaf-wise tree growth.
  61. 61
    Explain Bagging.
  62. 62
    Explain Boosting.
  63. 63
    Which reduces bias and which reduces variance?
  64. 64
    What is Gini Index?
  65. 65
    What is Information Gain?
  66. 66
    What is SMOTE?
  67. 67
    Explain ROC Curve.
  68. 68
    Explain Confusion Matrix.
  69. 69
    What is hyperparameter tuning?
  70. 70
    How does GridSearchCV work?
  71. 71
    Why did you choose only those parameters/solvers for GridSearchCV?
  72. 72
    How does Pipeline work?
  73. 73
    What is overfitting?
  74. 74
    How can you reduce overfitting?
  75. 75
    How did you handle outliers?
  76. 76
    How did you improve your model score?
  77. 77
    How can you further improve model performance?
  78. 78
    Why did you use R² score instead of Accuracy?
  79. 79
    Explain ANOVA Test.
  80. 80
    Explain Chi-Square Test.
  81. 81
    What is PCA?
  82. 82
    Explain the working of PCA.
  83. 83
    What is multicollinearity?
  84. 84
    How does Multiple Linear Regression work?
  85. 85
    What is RFE?
  86. 86
    How does RFE work?
  87. 87
    How do you select important features from a dataset?
  88. 88
    What feature extraction techniques have you used and why?
  89. 89
    What are soft class predictions and hard class predictions?
  90. 90
    Coding: Implement RFE to find the top 2 features.
  91. 91
    Coding: Load the Breast Cancer dataset and find the top features using RFE.
  92. 92
    Coding: Load the Diabetes dataset and find the top features using RFE.
  93. 93
    Coding: Load the Iris dataset and use SelectKBest with Chi-Square to find the top 2 features.
  94. 94
    Coding: Load the Iris dataset, train a Logistic Regression model with GridSearchCV, and print the classification report.
  95. 95
    Coding: Generate a word cloud for each label/class separately.
level2_103 MLP Level 1 Aug 5, 2026
  1. 1
    Explain your Kaggle/Colab notebook (project).
  2. 2
    What steps did you take to improve your model/score?
  3. 3
    Why did you use this model?
  4. 4
    Why did you use this imputer?
  5. 5
    Why did you use this scaler?
  6. 6
    Explain the working of the models you used.
  7. 7
    Explain the parameters/hyperparameters used in your model.
  8. 8
    What is a decision tree?
  9. 9
    Compare K-Means Clustering and KNN.
  10. 10
    What is a loss function?
  11. 11
    Why do we need a loss function?
  12. 12
    What is feature selection?
  13. 13
    Why do we need feature selection?
  14. 14
    What is data preprocessing?
  15. 15
    What is EDA (Exploratory Data Analysis)?
  16. 16
    What is dimensionality reduction?
  17. 17
    Explain PCA.
  18. 18
    What is hyperparameter tuning?
  19. 19
    Explain Logistic Regression.
  20. 20
    What is overfitting?
  21. 21
    What is underfitting?
  22. 22
    How can you prevent overfitting and underfitting?
  23. 23
    What is GridSearchCV?
  24. 24
    What is RandomizedSearchCV?
  25. 25
    What is Cross Validation (CV)?
  26. 26
    What are evaluation metrics?
lvl2_72 MLP Level 1 Aug 5, 2026
  1. 1
    What is stratify in train_test_split?
  2. 2
    Why do we use stratify?
  3. 3
    How can you find correlation between categorical variables?
  4. 4
    How can you determine whether a model is biased?
  5. 5
    These were already included in your master list: Theory about the models used. Precision, Recall, F1 Score. Suggestions for improving the model. Where else the model can be applied. General conceptual questions from the notebook.
  6. 6
    What is stratify?
  7. 7
    Why is stratify used?
  8. 8
    How do you measure correlation between categorical variables?
  9. 9
    How do you identify whether a machine learning model is biased?
mlp_level1_viva3 MLP Level 1 Aug 5, 2026
  1. 1
    Why did you choose this preprocessing strategy instead of another?
  2. 2
    Did you consider any alternative preprocessing techniques?
  3. 3
    Did you consider any alternative preprocessing techniques?
  4. 4
    If you had more time, what would you do to improve your model/project?
  5. 5
    Did you refer to any research papers, books, or other external resources while building the project?
  6. 6
    Show your Kaggle competition submissions.
  7. 7
    Create a preprocessing pipeline using: Mean imputation for numerical columns Mode (most frequent) imputation for categorical columns.
  8. 8
    Modify an existing pipeline to replace Median imputation with Mode.
  9. 9
    Fill categorical columns using SimpleImputer(strategy="most_frequent").
  10. 10
    Fill categorical columns with mode using Pandas (without SimpleImputer).
  11. 11
    Import the California Housing dataset.
  12. 12
    Perform an 80:20 train-test split.
  13. 13
    Load the Iris dataset and train an SGDClassifier.
  14. 14
    Load the Iris dataset and train SGDClassifier without train_test_split.
  15. 15
    Filter rows where: RegionCode == "Florida" AND TargetValue > 50000
  16. 16
    Why did you choose this preprocessing strategy?
  17. 17
    Did you consider alternative preprocessing methods?
  18. 18
    If you had more time, how would you improve the model/project?
  19. 19
    Did you use research papers/books/external resources?
  20. 20
    Show your Kaggle competition submissions.
  21. 21
    Create preprocessing pipeline with mean (numerical) and mode (categorical) imputation.
  22. 22
    Replace median imputation with mode imputation.
  23. 23
    Fill categorical columns using SimpleImputer(strategy="most_frequent").
  24. 24
    Fill categorical columns using Pandas mode.
  25. 25
    Load California Housing dataset and perform an 80:20 train-test split.
  26. 26
    Train an SGDClassifier on the Iris dataset (with/without train-test split).
  27. 27
    Filter a DataFrame where RegionCode == "Florida" and TargetValue > 50000.
level1_viva8 MLP Level 1 Aug 5, 2026
  1. 1
    What are the parameters of Logistic Regression?
  2. 2
    Can Logistic Regression be used to predict a continuous value (e.g., rainfall)? Why or why not?
  3. 3
    What are the parameters of Random Forest?
  4. 4
    What is the loss function of Random Forest?
  5. 5
    Difference between parameters and hyperparameters.
  6. 6
    What are the important XGBoost parameters?
  7. 7
    If XGBoost is overfitting, should you increase or decrease n_estimators? Why?
  8. 8
    If XGBoost is overfitting, should you increase or decrease learning_rate? Why?
  9. 9
    What does colsample_bytree / colsample_bylevel / colsample_bynode (referred to as col_by_sample) do?
  10. 10
    What is n_iter in RandomizedSearchCV?
  11. 11
    What does verbose mean?
  12. 12
    What does n_jobs do?
  13. 13
    What is the mathematical range of R² score?
  14. 14
    Why can R² be negative?
  15. 15
    What is data leakage?
  16. 16
    How did you ensure data leakage was avoided?
  17. 17
    Why should you not train the final model on the entire dataset before evaluation?
  18. 18
    How do you ensure reproducibility of results?
  19. 19
    What is autocorrelation?
  20. 20
    Explain the parameters of TF-IDF.
  21. 21
    Create a preprocessing Pipeline using: SimpleImputer RobustScaler KNNImputer OneHotEncoder
  22. 22
    Add RobustScaler to an existing preprocessing pipeline.
  23. 23
    Explain what changes after adding RobustScaler.
  24. 24
    Load the Diabetes dataset.
  25. 25
    Convert it into a Pandas DataFrame.
  26. 26
    Print the number of features.
  27. 27
    Fit a Logistic Regression model on it.
  28. 28
    Explain EDA overall, not line by line.
  29. 29
    Mention one important observation from the dataset.
  30. 30
    Show all model score comparisons.
  31. 31
    How many models were trained and how were they compared?
  32. 32
    Explain the mandatory Level-1 notebook sections one by one.
  33. 33
    Were any custom/user-defined functions used?
  34. 34
    Were any LLMs used while building the notebook?
  35. 35
    Parameters of Logistic Regression
  36. 36
    Parameters of Random Forest
  37. 37
    Loss function of Random Forest
  38. 38
    Difference between Parameters and Hyperparameters
  39. 39
    Important XGBoost parameters
  40. 40
    Effect of n_estimators on overfitting
  41. 41
    Effect of learning_rate on overfitting
  42. 42
    colsample_* parameter in XGBoost
  43. 43
    verbose
  44. 44
    n_jobs
  45. 45
    Mathematical range of R²
  46. 46
    Why R² can be negative
  47. 47
    Data Leakage
  48. 48
    How to prevent Data Leakage
  49. 49
    Why not train the final model on the full dataset before evaluation
  50. 50
    Reproducibility of results
  51. 51
    Autocorrelation
  52. 52
    TF-IDF parameters
  53. 53
    Can Logistic Regression predict rainfall? Why/Why not?
  54. 54
    Create/Edit preprocessing pipeline with RobustScaler, KNNImputer, OHE
  55. 55
    Load Diabetes dataset → DataFrame → Number of features → Fit Logistic Regression
  56. 56
    Explain one key insight from EDA
  57. 57
    Compare all trained models
  58. 58
    User-defined functions
  59. 59
    LLM usage in the notebook
level2_71 MLP Level 1 Aug 5, 2026
  1. 1
    Why did you use One-Hot Encoding instead of Label Encoding?
  2. 2
    For every preprocessing step: Why this technique and not another one?
  3. 3
    Why did you choose each hyperparameter in your model?
  4. 4
    Why did you use eval_metric='logloss'?
  5. 5
    Why did you choose a particular threshold value?
  6. 6
    Is Pickup Location ID useful? Why?
  7. 7
    Is Drop Location ID useful? Why?
  8. 8
    What do you observe about the target variable?
  9. 9
    Why do you say the target variable is normally distributed?
  10. 10
    Explain the advantages and disadvantages of each model you used compared to the others.
  11. 11
    Show graphs during EDA.
  12. 12
    Explain every graph and why you plotted it.
  13. 13
    Plot a graph comparing y_pred vs y_test (or prediction error) for every model.
Level3_9 MLP Level 1 Aug 5, 2026
  1. 1
    What is Support in a Classification Report?
  2. 2
    What is ROC-AUC? (Full form)
  3. 3
    How do you interpret the ROC-AUC score?
  4. 4
    What is Correlation?
  5. 5
    What do you infer from a Correlation Matrix?
  6. 6
    Explain Decision Tree working in depth.
  7. 7
    What are the assumptions of Linear Regression?
  8. 8
    What is the Curse of Dimensionality?
  9. 9
    How does Dimensionality Reduction work?
  10. 10
    Does dimensionality reduction remove/discard data points?
  11. 11
    What is random_state=42?
  12. 12
    Why is 42 commonly used?
  13. 13
    Why choose a learning rate like 0.01 / 0.1 / 0.2 instead of 1 / 2 / 3?
  14. 14
    What is the objective of your project?
  15. 15
    What is a Loss Function?
  16. 16
    What properties should a good loss function have?
  17. 17
    Explain Bias vs Variance.
LEVEL2_106 MLP Level 1 Aug 5, 2026
  1. 1
    Self introduction.
  2. 2
    Explain the problem statement.
  3. 3
    Explain your notebook to a non-technical person (e.g., your boss).
  4. 4
    Abstract questions about the dataset (not technical).
  5. 5
    Questions about college life.
  6. 6
    How many models did you try?
  7. 7
    How many models did you perform hyperparameter tuning on?
  8. 8
    Why didn't you submit earlier?
  9. 9
    Why F1 Score instead of Accuracy?
  10. 10
    How is F1 Score calculated?
  11. 11
    Macro vs Micro averaging.
  12. 12
    Why only Accuracy was considered?
  13. 13
    Did Hyperparameter Tuning improve your score?
  14. 14
    Which is your baseline model?
  15. 15
    What is imputation?
  16. 16
    Why did you choose this imputation strategy?
  17. 17
    What is scaling?
  18. 18
    What is normalization?
  19. 19
    What effect does normalization have on models?
  20. 20
    What do you think about outliers in your dataset?
  21. 21
    How did you handle class imbalance?
  22. 22
    Explain Hyperparameter Tuning.
  23. 23
    What did you do during Hyperparameter Tuning?
  24. 24
    Why didn't you perform Hyperparameter Tuning? (if missing from notebook)
  25. 25
    How much time did Hyperparameter Tuning take?
  26. 26
    Loss function of Logistic Regression.
  27. 27
    Loss function of Linear Regression.
  28. 28
    Overfitting vs Underfitting.
  29. 29
    How do you identify overfitting?
  30. 30
    How do you overcome overfitting?
  31. 31
    What is Cross Validation?
  32. 32
    Pipeline.
  33. 33
    PCA.
  34. 34
    Regularization.
  35. 35
    L1 vs L2 Penalty.
  36. 36
    Explain how L1/L2 Regularization works.
  37. 37
    Other ensemble models.
  38. 38
    Boosting vs XGBoost vs LightGBM.
  39. 39
    Explain how XGBoost works.
  40. 40
    Explain how LightGBM works.
  41. 41
    Can Gradient Boosting be converted into LightGBM?
  42. 42
    Which was your best model?
  43. 43
    Explain your best model in detail.
  44. 44
    Advantages of your best model.
  45. 45
    Difference between hstack() and vertical stacking (vstack()).
  46. 46
    Difference between Seaborn and Matplotlib.
  47. 47
    Load the Iris dataset.
  48. 48
    Train a Logistic Regression model on it.
  49. 49
    Self introduction
  50. 50
    Problem statement explanation
  51. 51
    Explain notebook to a non-technical person
  52. 52
    Abstract questions on dataset
  53. 53
    Questions about college life
  54. 54
    How many models tried?
  55. 55
    How many hyperparameter tuned?
  56. 56
    Why didn't you submit earlier?
  57. 57
    Why F1 instead of Accuracy?
  58. 58
    F1 Score calculation
  59. 59
    Macro vs Micro
  60. 60
    Imputation
  61. 61
    Imputation strategy
  62. 62
    Scaling
  63. 63
    Normalization and its effect
  64. 64
    Outliers
  65. 65
    Class imbalance handling
  66. 66
    Hyperparameter tuning explanation
  67. 67
    HPT improvement in score
  68. 68
    HPT execution time
  69. 69
    Logistic Regression loss function
  70. 70
    Linear Regression loss function
  71. 71
    Cross Validation
  72. 72
    PCA
  73. 73
    Pipeline
  74. 74
    Regularization
  75. 75
    L1 vs L2
  76. 76
    Overfitting vs Underfitting
  77. 77
    Detecting overfitting
  78. 78
    Preventing overfitting
  79. 79
    Boosting vs XGBoost vs LightGBM
  80. 80
    LightGBM working
  81. 81
    XGBoost working
  82. 82
    Other ensemble models
  83. 83
    Gradient Boosting → LightGBM conversion
  84. 84
    Best model explanation
  85. 85
    Advantages of best model
  86. 86
    hstack vs vstack
  87. 87
    Seaborn vs Matplotlib
  88. 88
    Logistic Regression coding on Iris dataset
Most Asked Topics

No tags yet.

View all →
Top Proctors
Created for educational purposes only. Questions are based on students' personal experiences and may not reflect actual exam content.