Augmented Wine Quality

The dataset utilized for this study is the Wine Quality dataset, which comprises 1,599 rows and 11 features related to the chemical properties of wine samples. The goal is to predict the ”quality” of the wine, a target variable that is an ordinal integer value, based on the following 10 features: fixed acidity, volatile acidity, citric acid, residual sugar, chlorides, free sulfur dioxide, density, pH, sulphates, and alcohol.

To test the model’s ability to identify and classify internal groupings within the dataset, we performed data augmentation. Specifically, we added Gaussian noise to the original feature values, effectively creating subgroups within the data. This noise was calculated as 10% of each feature’s mean and standard deviation, and was added to the data points to generate a new dataset. The result was a dataset that doubled in size to 3,198 rows, simulating internal group structures without providing explicit feature-based indications of these subgroups