This project demonstrates practical skills in statistical modelling, predictive analytics, and machine learning.
The work is divided into three main parts:

• Built a multiple linear regression model to predict compressive strength from mixture components.
• Applied Bonferroni correction for statistical significance testing.
• Used stepwise selection with BIC to identify the best subset of predictors.
• Predicted compressive strength for a new concrete mix and compared it to the existing industry standard.

• Fitted a decision tree using cross-validation to identify key predictors of heart disease.
• Built a logistic regression model with stepwise selection (BIC).
• Compared both models on test data using custom prediction statistics.
• Calculated patient-specific odds and used bootstrap confidence intervals to assess differences between individuals.

• Implemented k-nearest neighbours (k-NN) regression to smooth noisy mass spectrometry data.
• Measured performance with mean squared error (MSE) and selected optimal k via cross-validation.
• Evaluated trade-offs between background noise smoothing and accurate peak detection.
• Applied bootstrap resampling to estimate confidence intervals for predicted intensities.

• Delivered interpretable and accurate predictive models across regression, classification, and smoothing tasks.
• Logistic regression proved more reliable than decision trees for heart disease prediction.
• k-NN successfully smoothed noisy mass spectrometry signals, with k choice critical to balancing accuracy vs. noise.
• Applied bootstrap methods to provide statistical confidence in model predictions.

• Languages & Libraries: R, glm, rpart, randomForest, glmnet, kknn, boot
• Techniques: Regression, Classification, Decision Trees, Logistic Regression, Cross-Validation, Stepwise Selection, k-NN, Bootstrapping
• Applications: Predictive modelling, medical diagnostics, materials science, signal smoothing

Nashmia Shakeel