Below table 1 description of the attributes table 1

Below Table 1 description of the attributes.
Table 1: Attribute description
Attributes Description
fixed acidity Fixed acids, numeric from 3.8 to 15.9
volatile acidity Volatile acids, numeric from 0.1 to 1.6
citric acid Citric acids, numeric from 0.0 to 1.7
residual sugar residual sugar, numeric from 0.6 to 65.8
chlorides
Chloride, numeric from 0.01 to 0.61
free sulfur dioxide Free sulfur dioxide, numeric: from 1 to 289
total sulfur dioxide
Total sulfur dioxide, numeric: from 6 to 440
density Density, numeric: from 0.987 to 1.039
pH
pH, numeric: from 2.7 to 4.0
sulfates Sulfates, numeric: from 0.2 to 2.0
alcohol Alcohol, numeric: from 8.0 to 14.9
quality Quality, numeric: from 0 to 10, the output target

10
4.2. Feature selection
Feature selection is the method of selection of the best subset of
features that will be used for classification (Fauzi et al., 2017). Most
of the feature selection method is divided into a filter and wrapper, the
filter uses the public features work individually from the learning
algorithm and the wrapper evaluates the features and chooses
attributes based on the estimation of the accuracy by using a search
algorithm and specific learning model (Onan and Korukoğlu, 2017).
In this study, for a better understanding of the features and to examines
the correlation between the features. The Pearson correlation
coefficient is calculated for each feature in Table 1, this shows the
pairwise person correlation coefficient P, which is calculated by using
the below formula (Dastmard, 2013).
𝑃
?,?
=
cov (X, Y)
𝜎𝑋, σY
Where the 𝜎 is the standard deviation of the features X and Y and cov
is the covariance. The range of the correlation coefficient from -1 to
1. Point 1 value implies linear equation is describes the correlation
between X and Y strong positive, which is all data points are lying on
a line for Y increases as X increases. Where point -1 value indicates
that strong negative correlations between data points. All data points
lie on a line in which Y decreases as X increases. And point 0 indicates
that there is an absence of correlation between the points (Dastmard,
2013).
4.3. HyperParametr tuning
The grid search is a basic method for hyperparameter tuning. Perform
an inclusive search on the hyperparameter set specified by the user.
Grid search is suitable for several hyperparameters with limited search
space. The grid search algorithm is straightforward with enough
resources, the most accurate prediction can be drawn and users can

11
always find the best combination (Joseph, 2018). Running grid search
in parallel is easy because each test is run separately without affected
by the time series. The results of one experiment are independent of
the results of other experiments. Computing resources can be allotted
in a very flexible way. In addition, grid search can accept a limited
sampling range, because too many settings are not suitable. In
practice, grid search is almost preferable only when the user has
enough knowledge with these hyperparameters to allow the definition
of a narrow search space, and it is not necessary to adjust more than
three hyperparameters simultaneously. Although other search
algorithms may have more useful features, grid search is still the most
widely used method due to its mathematical simplicity (Yu and Zhu,
2020).