Authors :
Dr. M.S. Chaudhari; Kiran A. Ande; Hitanshu Shahare; Vaishnavi Helwatkar; Sejal Shinde; Divya Janbandhu; Sahil Rangari
Volume/Issue :
Volume 8 - 2023, Issue 12 - December
Google Scholar :
https://tinyurl.com/vtktb2y7
Scribd :
https://tinyurl.com/mr3tr8z2
DOI :
https://doi.org/10.5281/zenodo.10393843
Abstract :
Wine is the most popularly consumed
beverage in the world and its values are considered
important in society. Wine quality is always important to
consumers. If the quality is not good, you have to do a
different procedure from the beginning, which is very
expensive. As technology has evolved, manufacturers
have relied on various devices to test during the
development stage. Thus, they can get a better idea
about the quality of the wine, which of course saves a lot
of money and time. Predicting wine quality through
machine learning involves using algorithms to analyze
various factors that contribute to wine quality.
Therefore, for our research, we used a dataset of the
Portuguese red wine grape variety “Vinho Verde” from
Kaggle, which has different input variables based on
physicochemical tests. We use several machine learning
algorithms including Logistic Regression, SVC, Random
Forest, K-Neighbor Classifier, and Decision Tree. We
trained the dataset on these selected models and
compared the accuracy and precision to select the best
machine-learning algorithm, and we found out that the
Random Forest algorithm gave the best result out of the
six models respectively. Thus, helping us to predict the
quality of the wine on a scale of 0-10, considering a set of
characteristics. In addition, through feature selection
process we observed that alcohol content greatly affects
the wine quality, which was calculated using Random
forest’s Feature Importance attribute. We will use ANN
to build the model. Furthermore, training the model on
an unbalanced database leads to underestimation,
especially for minority classes. Therefore, we used
SMOTE to oversample the minority class in the target
variable. Our research explores the potential of these key
machine learning techniques to effectively predict wine
quality, providing insights for wine enthusiasts and the
wine industry to improve the selection and production of
quality wines.
Keywords :
Quality of Wine, Machine Learning, Random Forest Classifier, Decision Tree, Neural Network, Accuracy, Precision, Recall, F-1 Score.
Wine is the most popularly consumed
beverage in the world and its values are considered
important in society. Wine quality is always important to
consumers. If the quality is not good, you have to do a
different procedure from the beginning, which is very
expensive. As technology has evolved, manufacturers
have relied on various devices to test during the
development stage. Thus, they can get a better idea
about the quality of the wine, which of course saves a lot
of money and time. Predicting wine quality through
machine learning involves using algorithms to analyze
various factors that contribute to wine quality.
Therefore, for our research, we used a dataset of the
Portuguese red wine grape variety “Vinho Verde” from
Kaggle, which has different input variables based on
physicochemical tests. We use several machine learning
algorithms including Logistic Regression, SVC, Random
Forest, K-Neighbor Classifier, and Decision Tree. We
trained the dataset on these selected models and
compared the accuracy and precision to select the best
machine-learning algorithm, and we found out that the
Random Forest algorithm gave the best result out of the
six models respectively. Thus, helping us to predict the
quality of the wine on a scale of 0-10, considering a set of
characteristics. In addition, through feature selection
process we observed that alcohol content greatly affects
the wine quality, which was calculated using Random
forest’s Feature Importance attribute. We will use ANN
to build the model. Furthermore, training the model on
an unbalanced database leads to underestimation,
especially for minority classes. Therefore, we used
SMOTE to oversample the minority class in the target
variable. Our research explores the potential of these key
machine learning techniques to effectively predict wine
quality, providing insights for wine enthusiasts and the
wine industry to improve the selection and production of
quality wines.
Keywords :
Quality of Wine, Machine Learning, Random Forest Classifier, Decision Tree, Neural Network, Accuracy, Precision, Recall, F-1 Score.
