Wine industries use Product Quality Certification to promote their products and become a concern for every individual who consumes any product. But it's not possible to ensure wine quality by experts with such a huge demand for the product as it will increase the cost. It allows building a model using machine learning techniques with a user interface which predicts the quality of the wine by selecting the important parameters.

2. CONTENT
▪ INTRODUCTION
▪ OBJECTIVE
▪ DATA DESCRIPTION
▪ PROPOSED METHODOLOGY
▪ MODEL DESCRIPTION
▫ ARCHITECTURAL MODEL
▫ CART
▫ RANDOM FOREST TREE
▫ KNN CLASSIFIERS
▪ ARCHITECTURAL MODEL
▫ ALGORITHM COMPUTATION
▪ APPLICATION
▪ REFRENCES
2

3. INTRODUCTION
• About wine:
• Wine is a beverage made from fermented grape and other fruit juices with lower amount
of alcohol content.
• Quality of wine is graded based on the taste of wine and vintage. This process is time
taking, costly and not efficient.
• A wine itself includes different parameters like fixed acidity, volatile acidity, citric acid,
residual sugar, chlorides, free sulphur dioxide, total sulphur dioxide, density, pH,
sulphates, alcohol and quality.
• Problem statement:
• In industries, understanding the demands of wine safety testing can be a complex task
for the laboratory with numerous analytes and residues to monitor.
• But, our application’s prediction, provide ideal solutions for the analysis of wine, which
will make this whole process efficient and cheaper with less human interaction.
3

4. OBJECTIVE
• Our main objective is to predict the wine quality using machine learning through Python
programming language
• A large dataset is considered and wine quality is modelled to analyse the quality of wine
through different parameters like fixed acidity, volatile acidity etc.
• All these parameters will be analysed through Machine Learning algorithms like random forest
classifier algorithm which will helps to rate the wine on scale 1 - 10 or bad - good.
• Output obtained would further be checked for correctness and model will be optimized
accordingly.
• It can support the wine expert evaluations and ultimately improve the production.
4

5. DATA DESCRIPTION
5
• The dataset contains chemical descriptions of 6499
Portuguese “Vinho Verde” wines.
• There are 4899 entries for white wine, and 1600 entries for
red wines.
• The source of the data is taken from the UCI Machine
Learning Repository, provided by Paulo Cortez, from the
University of Minho, Portugal.
5

6. DATA
DESCRIPTION
6
Attributes Description
pH To measure ripeness
Density Density in gram per cm3
Alcohol Volume of alcohol in %
Fixed Acidity Impart sourness and resist microbial infection, measured in no. of
grams of tartaric acid per dm3
Volatile Acidity no. of grams of acetic acid per dm3 of wine
Citric Acid no. of grams of citric acid per dm3 of wine
Residual Sugar Remaining sugar after fermentation stops
Chlorides no. of grams of sodium chloride per dm3 of wine
Free Sulfur
dioxide
no. of grams of free sulphites per dm3 of wine
Total Sulfur
dioxide
no. of grams of total sulfite (free sulphite+ bound)
Sulphates no. of grams of potassium sulphate per dm3 of wine
Quality Target variable, 1-10 value6

7. CONCLUSION OF DATA ANALYSIS
FIRST
The two most important features among
all 12 attributes are Sulphur dioxide (both
free and total) and Alcohol.
LAST
Volatile acidity contributes to
acidic tastes and have negative
correlation to wine quality.
SECOND
The most important factor to decide the quality of wine is alcohol,
higher concentration of alcohol leads to better quality of wine and
lower density of wine.
7

8. PROPOSED METHODOLOGY
▪ It gives insights of the dependency of target variables on independent variables using
machine learning techniques to determine the quality of wine because it gives the best
outcome for the assurance of quality of wine.
▪ The dependent variable is “quality rating” whereas other variables i.e. alcohol,
sulphur etc. are assumed to be predictors or independent variables.
▪ While hindering the effectiveness of the data model, various types of errors have
occurred like over fitting, introduced from having too large of a training set and bias
occur due to too small of a test set.
8

9. MODEL DESCRIPTION
9

10. “
10

11. CLASSIFICATION AND REGRESSION TREE
(CART)
▪ CART is a decision tree used for analysing both datasets (red and white wine).
▪ The decision trees produced by CART are always binary, containing exactly two
branches for each decision node.
▪ The CART algorithm grows the tree by conducting for each decision node, an
exhaustive search of all available variables. All possible splitting values, selecting the
optimal split
11

12. RANDOM FOREST
▪ Random forest is a method of classification, regression and other tasks, that operate
by constructing a multitude of decision trees at training time and outputting the class
that is the mode of the classes (classification) or mean prediction (regression) of the
individual trees.
▪ Following are some of the features of random forest algorithm:
1. It runs efficiently on large databases.
2. It gives estimates of what variables are important in the classification.
3. It generates an internal unbiased estimate of generalization error as the forest
building progresses occur.
12

13. “
13

14. K-Nearest-Neighbourhood Classifiers
▪ This classifier technique is depended on learning by analogy; this means a comparison
between a test tuple with similar training tuples.
▪ The training tuples are described by n attributes. Each tuple corresponds a point in an n-
dimensional space. All the training tuples are stocked in an n-dimensional pattern space. For
an unknown tuple, a k-nearest-neighbourhood classifier searches the pattern space for the k
training tuples that are closest to the unknown tuple. k training tuples are called as the k
“nearest neighbours” of the unknown tuple.
▪ “Closeness” is a metric distance, likewise Euclidean distance between two points or tuples,
say, 𝑋1 = (𝑥11, 𝑥12……….. 𝑥1𝑛) and 𝑋2 = (𝑥21, 𝑥22……….. 𝑥2𝑛), is:
dist (X1, X2) = 𝑖−1
𝑛
𝑥1𝑖 − 𝑥2𝑖
2
14

15. ARCHITECTURE MODEL
A flow chart that briefly describes processing of our
software application.
15

16. 16
16

17. “
Enter all 12 data elements of newly developed wine.
• Entered wine data is further divided into two sets (white wine datasets or red wine datasets) based on alcohol value.
• Based on respective datasets a binary tree is formed ( using random forests algorithm)
• Random forest algorithm
• recode(quality([3,4,5]=0,[6,7,8]=1))
• RandomForestClassifier(estimator=25)
• return prediction, confusion matrix, accuracy
• Total value is evaluated which is root value (using KNN)
• knn
• classify(X,Y,x)
• for i=1 to m do
• compute distance d(X,x)
• compute set I containing indices for the k smallestdistance d(X,x)
• cluster(X)
• return majority label for {Y,where i E I}
• Evaluated root value is actually our required parameter value i.e. depended variable or quality of wine (on scale of 1 to 10).
• Later on this estimated value is compared with datasets value and our algorithm is trained our defined datasets so as to for
correctness and model will be optimized accordingly.
ALGORITHM COMPUTATION
17

18. APPLICATION
▪ Results will be used by the wine
manufacturers to improve the quality of the
future wines.
▪ Certification bodies can also use the result
for quality control.
▪ Results can be used to make wine selection
guides for wine magazines.
▪ Results can be used by consumers for wine
selection
18

19. REFRENCES
▪ [1] Y. Er, “The Classification of White Wine and Red Wine According to Their Physicochemical
Qualities,” Int. J. Intell. Syst. Appl. Eng., vol. 4, no. SpecialIssue-1, pp. 23–26, 2016.
▪ [2] E. Summary, W. P. Monitoring, W. Quality, W. Safety, and W. Complexity, “Wine Analysis :
from ‘Grape to Glass’ An analytical testing digest of the wine manufacturing process,” 2016.
▪ [3] A. Ghosh, “Project Report : -Red Wine Quality Analysis Final 3 . An empirical Red Wine
Quality Analysis of the Portuguese ‘ Vinho Verde ’ wine,” no. December 2017, 2018.
▪ [4] Y. Gupta, “Selection of important features and predicting wine quality using machine
learning techniques,” Procedia Comput. Sci., vol. 125, pp. 305–312, 2018.
▪ [5] P. Model, L. Regression, and R. Studio, “Building and Evaluating a Predictive Model w/
Linear Regression in RapidMiner Studio,” 2018.
▪ [6] B. Tajini and O. C. Paris, “Badr Tajini – On Campus Paris – DSTI 2017,” vol. 47, no. 4, pp.
547–553, 2017.
19
19

20. 20 20