Wine Quality Prediction

01 - The problem

Raw quality data was riddled with missing values and outliers that quietly wrecked naive models.

02 - Approach

Ran a full preprocessing pass - missing-value imputation, outlier treatment, scaling - then benchmarked Random Forest against a Support Vector Classifier to see which handled the cleaned feature space better.

03 - Outcome

Demonstrated how much of model performance is won in preprocessing, with a documented comparison between ensemble and margin-based approaches.