The report focuses on data mining approach to predict human wine taste preferences. A large data set is considered with white and red wine samples (“Vinho Verde” wine from Portugal). The inputs include objective tests (e.g. PH values) and the output is based on sensory data (median of at least 3 evaluations made by wine experts). Each expert graded the wine quality between 0 (very bad) and 10 (very excellent). Due to privacy and logistic issues, only physicochemical (inputs) and sensory (the output) variables are available (e.g. there is no data about grape types, wine brand, wine selling price, etc.).
Each record contains 12 attributes. Each record contains a set of attributes and one attribute …show more content…
A relative absolute error of 83.48% indicates that the forecasts are 83.48% similar to the eventual outcomes.
The number in the output indicates wine quality. The wines are rated from 1-10. So the numbers also fall in the same range. The grades taken here are given by expert opinion. This forms the training data set. This is used to build model. Once the model is built, the same data is used to validate the model as shown below.
Log obtained from WEKA for cross validation
== Run information ===
Scheme:weka.classifiers.functions.MultilayerPerceptron -L 0.3 -M 0.2