White Wine Quality Analysis with Regression and Model Trees in R

In this R tutorial, we will be estimating the quality of wines with regression trees and model trees. Machine learning has been used to discover key differences in the chemical composition of wines from different regions or to identify the chemical factors that lead a wine to taste sweeter. In most cases, wine experts rate wine that can predict whether the wine is labeled as the bottom or top shelf.

This R tutorial will be in remembrance of my wife’s grandmother as she loved to have a glass of white wine (maybe two or three) with friends and family. Good times with white wine!

The methods used will be regression trees and model trees to create a system capable of mimicking ratings of wine. This will allow the winemakers to identify the key factors that contribute to better-rated wines.

Install and Load Packages

Below are the packages and libraries that we will need to load to complete this tutorial:

Download and Load the White Wine Dataset

Since we will be using the wine datasets, you will need to download the datasets. The datasets are already packaged and available for an easy download from the dataset page or directly from here White Wine – whitewines.csv

Example import command for the red and white wine excel CSV file:

View the White Wine Dataset

str() function

Histogram of the Quality of Wine

Create Wine Train and Test Models

Methods for training a model on the data

The rpart() will be used to specify quality as the outcome variable and use the dot notation to allow all the other columns in the wine_train data frame to be used in predictors.

Decision Tree Visualization

As one can see from the below, the visualization in the decision tree is much easier to read. Also, the digits parameter rounds all digits to the 3 places.

fallen.leaves() addition to the decision tree

This addition will show visualizations with the dissemination of regression tree results, as they are readily understood even without a mathematics background. The lead nodes are predicted values for the examples reaching that node.

Test Data Prediction

We must now make predictions on the test data, we use the predict() function. This will return the estimated numeric value for the outcome variable.

Correlation

We can now check the correlation between the predicted and actual quality values provides a simple way to gauge the model’s performance. As one can see that the outcome is 0.54, which is acceptable but not ideal. This correlation only measures how strong the predictions are related to the true value. This is not a measure of how far off the predictions were from the true values.

Mean Absolute Error

Let’s turn the tables and thinking of another way we could improve the model’s performance. We could consider how far, on average, its prediction was from the true value.

As shown above, on average, the difference between our model’s predictions and the true quality score was about 0.59.

The above shows room improvement. As one can see, MAE shows 0.59, which comes closer to the average to the true quality score than the imputed mean, MAE 0.67.

M5 Algorithm Improvement

The M5 algorithm will return a model tree object that can be used to make predictions. Also, we will use the predict function that will return a vector of predicted numeric values.

Let’s predict the unseen data and the correlation is rather high.

Decision trees were used for numeric prediction to model the wine data. The model trees, which builds a regression model at each leaf node in a hybrid approach. However, the latest cor() did not improve much, it did surpass the performance of the neural network model published. Our model output was close to the published mean absolute error value of 0.45 for the support vector machine model. This was a much simpler learning method for the data used.