PREDICTIVE ANALYTICS B.Tech. IV Year I Sem JNTUH R-18

• Derive the normal equations for least squares regression and explain their significance.

• Discuss the bias-variance trade-off in the context of linear regression and how to achieve a balance.

• Compare and contrast ridge and lasso regression, and when would you choose one over the other?

• Explain the difference between linear discriminant analysis and logistic regression, and their suitability for different tasks.

• Implement the perceptron learning algorithm for binary classification and explain the convergence criteria.

Previous Year Paper Questions:

• A dataset contains house prices and features. Describe how you would use linear regression to predict prices and the potential challenges.

• A bank wants to classify loan applicants as likely to default. Build a logistic regression model and evaluate its performance.

• Compare the performance of ridge and lasso regression on a real-world dataset (e.g., spam classification, stock prices).

Unit II: Model Assessment and Selection

• Explain the bias-variance trade-off and its implications for model selection.

• How does cross-validation help estimate the generalization error of a model? Discuss different techniques (e.g., k-fold).

• Explain the advantages and disadvantages of bootstrap methods for model assessment.

• Compare and contrast the Bayesian approach and the frequentist approach to model selection.

• Describe how the optimism of the training error rate can lead to overfitting and how to avoid it.

Previous Year Paper Questions:

• You have two models with differing training errors. Explain how cross-validation helps choose the model likely to perform better on unseen data.

• A company wants to predict customer churn. Describe how you would use the Bayesian approach to select the best model among several candidates.

• Discuss challenges of training decision trees and boosting algorithms and how to overcome them.

Unit III: Additive Models, Trees, and Boosting

• Explain the concept of an additive model and how it can be used for both regression and classification.

• How do decision trees work for classification, and what are the advantages of using them?

• Describe the AdaBoost algorithm and explain how it combines weak learners to create a strong learner.

• Give examples of real-world applications where additive models, trees, and boosting are commonly used.

• Explain gradient boosting and its applications in numerical optimization problems.

Previous Year Paper Questions:

• Build a decision tree model to classify emails as spam or not spam and analyze its performance using appropriate metrics.

• Compare and contrast the performance of AdaBoost with a single decision tree on a given classification task.

• Explain how gradient boosting can be used for numerical optimization problems and provide an example application.

Unit IV: Neural Networks (NN), Support Vector Machines (SVM), and K-nearest Neighbor

• Explain the basic architecture of a feedforward neural network and how it learns through backpropagation.

• Discuss the challenges of training neural networks and techniques to overcome them (e.g., overfitting, vanishing gradients).

• How do support vector machines work for classification? What are their advantages over other methods?

• Explain the k-nearest neighbor algorithm and its applications in both classification and regression.

• Give examples of real-world applications where neural networks, SVMs, and k-nearest neighbors are commonly used.

Previous Year Paper Questions:

• Build a simple neural network to classify images and evaluate its performance.

• Compare and contrast the performance of SVMs with logistic regression for a text classification task.

• Explain how k-nearest neighbors can be used for anomaly detection and provide an example application.

Unit V: Unsupervised Learning and Random Forests

• Explain how association rule mining can be used to identify patterns in market basket data.

• Describe how k-means clustering can be used to group customers into different segments.

• Explain how principal component analysis can be used to reduce the dimensionality of a dataset.

• How can random forests be used for both regression and classification tasks? What are their advantages?