Content uploaded by Mahdi Jasim
Author content
All content in this area was uploaded by Mahdi Jasim on Sep 30, 2022
Content may be subject to copyright.
How to Cite:
Jasim, M. N., & Hamid, A. B. (2022). Food recommendation system based on nutritional
needs of human beings and user preferences. International Journal of Health
Sciences, 6(S4), 4025–4038. https://doi.org/10.53730/ijhs.v6nS4.9031
International Journal of Health Sciences ISSN 2550-6978 E-ISSN 2550-696X © 2022.
Manuscript submitted: 27 March 2022, Manuscript revised: 18 May 2022, Accepted for publication: 9 June 2022
4025
Food recommendation system based on
nutritional needs of human beings and user
preferences
Mahdi Nsaif Jasim
PhD, University of Information Technology and Communications, Iraq
Ahmed Bahaddin Hamid
MBCHB, CABMS/CM, Baghdad Medical City, Iraq
Abstract---Introduction: Nowadays the food types became so diverse
and complicated, so human needs+ professional assistance to make
his best choices especially after foods became global parameter. Food
Recommendation System is a smart system that provides the best
suggestions to the beneficiaries to know the best choices to their
needs. Moreover, the human activities and lifestyle are affected by
another types of dietaries in other foods. There is need for everybody
to know what the nutrition is he/she needs. So, this research
responding to these needs. The goal of the proposed system is to
propose a system that provides recommendations for foods that are
rich in nutritional components that people need in their daily lives
based on computational model and expert preferences. Objective: The
research aims to design and implement a food recommendation
system has the ability to coordinate both user preferences and data
clustering techniques to produce high accuracy recommendations.
Material and methods: The proposed method focuses on merging
computational model and user preferences to give the user the best
recommended list of food options. Clustering techniques are
approaches used in Recommendation system applications to group
different foods according to the similarities in nutrition values. The
proposed solution uses a combination of Silhouette algorithm and the
K-means for weights to obtain the optimum number of clusters early
before the clustering process. The weight of each cluster is calculated
to use it in determining the rich cluster of food ingredients among
other clusters. Results: The results showed an accuracy of 96%.
Precision calculation performed using Silhouette algorithm.
Conclusions and Recommendations: The system knowledge is limited
to the training dataset used. The system performance is upgraded if
we use dataset with more food data. The system provides consultation
results affected by user preferences.
4026
Keywords---recommendation systems, dietary systems, clustering
systems, K-means, Silhouette algorithm, food recommendation.
Introduction
Healthy people usually choose healthy food carefully from a large number of foods
that contain vitamins and mineral elements, which are correspond to the needs of
each person [1]. So, recommendation systems became more and more popular in
various areas of our life. Where recommendation systems can provide great
benefit when its necessary to recommend items of food to the people that meet
their needs [1]. In this paper, foods recommendation relying on machine learning
using the clustering approach. The recommendation system known as the smart
system that suggests the best suitable target result based on the analysis of user
information and the analysis of user interests. Several studies have proven that
the recommendation system is the powerful way to solve many problems [1].
Recommendation systems suggests items based on users' past behavior,
preferences, and personal data. Given the diversity of data, diversity of
information, and wide range of products, recommendation systems are essential
to provide recommendations for products and other elements [2].
Recommendation systems help users to find the best option from a huge number
of possible options as it meets the requirements of users in a very short time
through knowing their initial preferences [3]. Recommendation systems have
achieved great performance in various fields such as music, movies, news, books,
and products. When user's preferences are unknown to the system, so this kind
of problem called the cold start problem [5] which in addition to vague of user
preferences could be overcome with the help of clustering methods [6]. The items
with the highest similarities grouped together in one group and the items with the
highest difference grouped into different groups [7,8]. Clustering operations
require a large set of data to achieve high accuracy in the element prediction
process and improve the clustering process [9]. To improve the accuracy of the
results in this proposed system, we combine a silhouette algorithm that used to
determine the best initial number of clusters. the clustering algorithm k-means
used to cluster the food types. in this paper, description to the proposed system,
food dataset, the proposed algorithm that works with a weights method, and how
to determine the clusters that fit to the user's needs are discussed.
Related Works
To investigate the recent updates and trends of using the clustering techniques in
the human food recommendation systems, the next is a short review to them. In
2020, A Recommender System for Healthy and Personalized Recipe
Recommendations was proposed by Florian Pecune, Lucile Callebert, and Stacy
Marsella proposed a recommender system for healthy food recommender system
used to change the user food consumption behavior [28].
In 2018, Rui Maia, and Joao C. Ferreira, proposed a food recommendation
system, using medical records and mobile devices. Users who pass by food places
receive recommendations for types of food based on the available foods that
suggested depending on whether these foods are suitable for everyone's health.
4027
Where a method used for recommending recipes based on feature engineering,
and matrix factor [14].
In 2017 D.S.Gaikwad et al: built a nutritional recommendation system using
data-mining methodology that creates relationships and patterns between foods.
The system uses advanced data mining techniques such as machine learning.
Through its interface, the system receives the information interactively, which in
turn provides the ideal solution for the end user [11].
In 2015, Mehdi Elahi and his et al., Design an interactive food system through a
set of interactions amongst short terms and long terms requirements of clients.
Long-term preferences taken by asking the user to classify generally desired
recipes and the user asked to find preferences for short-term recipes, to
determine which ingredients they want in the recipe to be prepared. Based on the
similarities between the two types of preferences [12].
In 2013 Sumedh Sawant and Gina Pai presented a dataset from which to extract
content-based collaborative features used to find restaurant and customer
profiles. Hybrid cascade used to K-nearest neighbor clustering, and a two-part
weighted graph projection is used. The system evaluated using Root metrics Mean
Squared Error and mean absolute error [13].
In 2010, Maiyaporn Phanich, Phathrajarin Pholkul, and Suphakant Phimoltares:
Dietary recommendation suggested using dietary data for diabetics. Alternative
nutrients recommend the use of the SOM and K-mean clustering algorithm to
analyze foods with similarity and suitability of eight beneficial nutrients important
to diabetics [10].
Problem Statement
People need many vitamins, nutrition, and supplements in their daily lives, many
people have their choice to buy these supplements, which are manufactured by
many food and drug companies. Because most of these nutritional supplements
are costly and unavailable in times of crisis and sometimes have side effects, that
makes natural sources of food the best solution to avoid these problems. The
proposed system greatly contributes helping people to find their nutritional
requirements of nutrition by suggesting foods rich with these components
through a mixture of algorithms, relying on the components of the materials to
achieve the desired goal in a list of foods under study.
The Objectives of Research
Design and implementation of food recommendation system to cope with huge
number of food sources, plants and recommends a set of them rich with vitamins
and minerals that people need as mails or food supplements making benefits of
user preferences.
4028
The Proposed Method
To build a dietary system recommends food rich with vitamin and minerals, there
is a need to collect big data about various types of food. And store the data in
database designed for this purpose, Figure 1 depicts the proposed system
architecture.
Figure. 1 The proposed system block diagram.
The Food Database
The data base contains a large collection of food (2000 item). Each type of food
contains a group of vitamins, iron, and nutritional components . Table 1
represents the structure database for each type of food.
Table 1. Information of food Database
Food ingredients
Measuring unit.
Carbohydrate
G
Manganese
Mg
Pantothenic Acid
Mg
Protein
G
Iron
mg
Magnesium
mg
Sodium
mg
Zinc
mg
4029
Vitamin A
Vitamin B6
mg
mg
Vitamin B12
mg
Vitamin C
mg
Vitamin E
mg
Vitamin K
mcg
Food Nutrition’s Weights
Calculating the weights of each food ingredients that contains vitamins, nutrition
and other components required to determine their importance to user. The user
preferences are adjustable through indicators located in the system user
interface, which enables the user to determine the importance value of all the
required nutritional components and the other are negligible by setting their
indicator to zero value. The weights of each type of food are calculate by equation
1 and the Figure 2 shows how the user determines the importance of each
feature.
……. (1)
Nutritional ingredient degrees of importance are user input as shown in figure 2.
Figure. 2: The proposed system user interface
4030
K-Means Clustering Algorithm
K-means is one of the oldest and popular clustering techniques. I’s
capable of dealing with large datasets [15,18]. This algorithm finds the
approximate solution. Some features of K-means are simplicity, easy to operate,
accommodation to data format, medium speed and scalable [16]. The distance is
used to measure the difference of the K clusters in the dataset. [17]. The dataset,
for example X, contains multiple data points. K is the initial number of clusters
which is required and usually specified by the user. By using the Euclidean
distance, the similarity extent is calculated between items and centroid of each
group [18]. This algorithm used very widely to determine common elements in
form of groups accurately separated and sensitive to modification [19].
Clustering algorithms collect primary information based on the attributes,
characteristics, and diversity of that information to perform the clustering process
by computing the least distance between the centroid of each cluster and the
current elements to put into any cluster [20]. The determination of optimal
number of clusters which is decided by the user is key weaknesses of K-means
algorithm. The random selection of the centroids affects both the accuracy of the
clustering results and the running time [21].
Algorithm 1: K-Means Algorithm [22]
Input: Set the amount of data.
K: Number of cluster (user input).
Output: Data grouped to k-number of clusters.
Start
1: determine the number of centroids k.
2: Repeat.
3: Construct K clusters by allocating each data point to the closest
centroid.
4: Find each cluster centroid, iteratively, until the data
movement stops.
End
Solving the K-Means Weakness
There are many methods used to solve this problem in a smart way by
determining the best number of clusters for each type of data, and one of these
methods is the silhouette algorithm [23,24].
Silhouette Algorithms
Silhouette used to measure clustering quality of each data point and how well the
clustering quality. Average items of the entire dataset or an individual set is a
measure of a clustering quality [25]. It combines decision and cohesion.
Consistency is the similarity measure amongst the object and the cluster
centroid. The comparison is made by the Silhouette algorithm, the result in the
range -1 to 1. If the Silhouette value is close to 1, it refers to a close match
between the current object and the cluster members while the -1 refers to poorly
close to the neighbor members. If a dataset is created by a model and it has a
4031
relatively high Silhouette value then the model is appropriate and acceptable
[26,27]. The silhouette value is computed as in equation (2)
………….......(2)
=
a(i) is the sample i intra-cluster dissimilarity.
b(i) is the sample i inter-cluster dissimilarity.
s(i) refers to Silhouette value.
Algorithm 2: Silhouette algorithm for determining K
Input: weight of each food in dataset
Output: number of cluster (K)
Start
Step 1: For i = 2 to i=50
Step 2: Intra_Mean = (weights of foods / sum of all weights of food in cluster)
Step 3: Inter_Mean (sum of all centroids of clusters) / i
Step 4: if Inter_Mean ≥ Intra_Mean then
Max inter_Mean
Else
Max Intra_Mean
End if
Step 5: Silhouette (i) = (Inter_Mean – Intra_Mean) /
max(Inter_Mean, Intra_Mean)
Step 6: End for
Step 7: optimal number of cluster (k) = max (Silhouette (i))
END
The Modified K-Means Algorithm
Model construction requires a modification to the clustering Algorithm which is
applied to cluster foods. The proposed method operates on the average weights, it
arranges the clusters from the highest nutritional and vitamin values to the
lowest. The work is divided to three stages. In stage 1(data pre-processing)
weights are considered. In stage 2, the average weight for each cluster is
calculated. In stage 3 threshold value is applied to determine the clusters whose
average value higher than the threshold (the winner clusters).
Algorithm 3: The modified K-Means Clustering Algorithm [16]
Inputs:
- Weights of foods calculated using Equation (1),
- Number of clusters (k) computed by algorithm (2),
- Threshold value.
Output: clusters that only contains the recommended foods.
Start
4032
1: Read weights.
2: Normalize data.
3: Make the weights the centroids of clusters (k) in random manner.
4: Compute the distance between each weight and all
Centroids, the distance is computed using the Euclidean distance.
5: Collect the weights to the nearest centroids.
6: Calculate new centroids for each cluster.
7: Repeat steps 3 through 5 until stability occurs.
8: Calculate the average-weight for each cluster by calculating the total weight of
the cluster divided by the number of elements in the cluster.
9: If average-weight >= the threshold value.
10: sort the clusters from the largest average-weight to the lowest average-
weight.
Step 11: End if
Step12: Display suggested clusters to the user
End
Average Cluster Weights computation
After applying the proposed K-means algorithm the number of clusters (k) is
determined and the data is clustered around the k-centroid. To compute the
average weight of each cluster, each cluster food members weights are summed
and the sums divided by the count of members of each cluster to produce the
average weight of the clusters.
Determine the Winning Clusters
The computed threshold is used to determining the best fit clusters which are the
clusters with high value of the nutrition subject to query. The threshold value is
compared with the average weights of each cluster. If the resulting average value
is greater than or equal to the threshold value, the cluster is recommended to the
user.
Results and Discussion
C# language, and SQL databases are the SW tools used to develop the proposed
system and the SQL-server database is used to store the input dataset and the
clustering. The proposed system is a mixture of both modified K-means and
Silhouette algorithms to enhance the clustering procedure and boost clustering
quality. The input dataset which describes the properties is presented to the
clustering procedure, and the centroid seeds (K) is computed by Silhouette
algorithm. Modified K-means accepted K as input. The modified procedure
calculates and displays the most fit clusters of food. To determine the weights of
each food in database, the user have to set what vitamins or nutritional
components are needed and what the degree of importance of each component.
The table 2 and table 3 shows how to find total weights.
Table 2: The importance of every vitamin and nutritional component
Nutritional ingredient
Importance degree
Carbohydrate
20%
4033
Manganese
20%
Pantothenic Acid
0%
Protein
20%
Iron
20%
Magnesium
0%
Sodium
0%
Zinc
0%
Vitamin A
0%
Vitamin B12
0%
Vitamin B6
0%
Vitamin C
0%
Vitamin E
0%
Vitamin K
20%
The user determining the importance of each nutritional ingredient. The weights
of the required component are calculated using Equation 1. Table(3) shows the
calculated weights of each type of food.
Table 3: Compute of the total weight
The results of clustering accuracy are done and the efficiency test is done. The
silhouette algorithm is exploited to fit the nutritional ingredient that belongs to
any group and to predict the optimal number of clusters for groups (as shown in
Fig. 3 and Table 4). This test clarifies the effect of computing the initial number of
clusters which is find by silhouette algorithm on clustering accuracy. Table 4
shows the difference of average value of Silhouette in accordance with number of
clusters in the range of 2 to 25.
Table 4: Average value of Silhouette for each number of cluster
Number of Clusters
Average value of Silhouette
K=2
K=3
K=4
K=5
K=6
K=7
K=8
K=9
K=10
0.542797072860022
0.527739315820888
0.530885551759057
0.531358297674783
0.547901661204242
0.560138526509069
0.557143213508186
0.542253717356613
0.560266138028575
Name of food
Carbohydrate
Manganese
Protein
Iron
vit K
Total
weight
Abiyuch,raw
17.6
0.182
1.5
1.61
0
4.1784
Alfalfa seeds,
2.1
0.188
3.99
0.96
30.5
7.5476
Allspice,grou
nd
72.12
2.943
6.09
7.06
0
17.6426
Amaranth
leaves,raw
4.02
0.885
2.46
2.32
114
0
229.937
4034
K=11
K=12
K=13
K=14
K=15
K=16
K=17
K=18
K=19
K=20
K=21
K=22
K=23
K=24
0.578540255902207
0.544860757451324
0.56413767677866
-0.597994921240362
-0.71501838741997
0.544606762744968
0.543688437448752
0.539926425366547
0.529759334158076
0.533250073806159
-0.728701252981128
0.542000926767654
-0.689600234433421
0.536557162439192
Fig. 3 The relation between computed-K and clustering results (k=11)
Table 5 shows the relation between the number of clusters, number of foods and
the cluster average weight.
Table 5: Number of foods and weight of each cluster
Cluster number
Number of foods
Cluster average weight
0
73
17.8023698630137
1
89
9.86706966292135
2
50
26.482716
3
379
1.94213034300792
4
13
72.42948
5
503
3.95976500994036
6
16
104.3768
7
341
5.947696
8
6
341.7024
9
9
172.9516
10
22
42.8313
4035
Fig. 4 The clusters average weights
After that, the clusters are arranged so that the clusters that have the highest
weights among the other clusters will be the best clusters and their weights are
greater than the pre-determined threshold value as in (Table 6).
Table 6: Food clusters recommended to the user
Clusters Number
Average weights of
each cluster
Threshold
Value
State of clusters
8
341.7024
100
Suggest to user
9
172.9516
100
Suggest to user
6
104.3768
100
Suggest to user
4
72.42948
100
Ignore
10
42.8313
100
Ignore
2
26.482716
100
Ignore
0
17.8023698630137
100
Ignore
1
9.86706966292135
100
Ignore
7
5.947696
100
Ignore
5
3.95976500994036
100
Ignore
3
1.94213034300792
100
Ignore
The results shown in Table 6 show that the clusters 8,9,6 only have average
weight higher than the threshold value 100 compared to other clusters.
Accordingly, cluster 8 contains the highest recommended foods according to the
user’s requirements.
Table 7: The recommended foods are in cluster 8
Type of foods
Name of food
Carbohydrate
Manganese
Protein
Iron
vit K
Spices
Spices,basil,dried
60.96
3.167
14.37
42
1714.
5
Spices
Spices,thyme,drie
d
63.94
7.867
9.11
123.6
1714.
5
Sage
Sage,ground
60.73
3.133
10.63
28.12
1714.
4036
Limitations of the Proposed System
There are two weaknesses points:
The system weakness is that it neglected a person’s medical history. A person
may have common diseases such as diabetes or hypertension etc. The system
may recommend food rich in ingredients that may cause harm to the patient. The
system recommends in types of food are rich in ingredients may be excess of the
actual need of the patient because it ignores age, gender and lifestyle (nutrition,
activity). Therefore, we have to enter the person’s information to get the right
items and the right amounts of each item that fit persons need.
Conclusion and Future Works
This paper deals with a recommendation system for foods suits a specific
nutritional ingredient needed by the clients. The system is tested on database
with 2000 items of food types, the system results showed that it’s more accurate
than traditional k-means algorithms. The results tested with one of the methods
for evaluating the clustering, using Silhouette algorithm to calculate the inter and
intra cluster and gave a value (0.578540255902207) with a small run time (13.52
seconds). Its suggested develop the current system to deal with ingredients and
recipes.
References
1. Bundasak, Supaporn. "A healthy food recommendation system by combining
clustering technology with the weighted slope one predictor." 2017
International Electrical Engineering Congress (iEECON). IEEE, 2017.
2. Maiyaporn Phanich, et al ” Food Recommendation System Using Clustering
Analysis for Diabetic Patients, Advanced Virtual and Intelligent Computing
(AVIC) Research Center Department of Mathematics, Faculty of Science,
Chulalongkorn University Pathumwan, Bangkok, Thailand, Proc. 6th Int’l
Conf Machine Learning and Cybernetics, pp. 2953-2957, August 2007
3. K. Vipin, An Extensive Survey of Clustering Methods for Data Mining.
http://www-users.cs.umn.edu/~han/dmclass/
4. Razak, Tajul Rosli, et al. "Career path recommendation system for UiTM
Perlis students using fuzzy logic." 2014 5th International Conference on
Intelligent and Advanced Systems (ICIAS). IEEE, 2014.
5. Sachin, J., et al. "Location Based Agricultural Product Recommendation
System Using Novel KNN Algorithm." (2019).
6. Naw Naw and Ei Ei Hlaing. "Relevant Words Extraction Method for
Recommendation System." Buletin Teknik Elektro dan Informatika Vol. 2, No.
3 (2013): 169-176.
5
Parsley
Parsley,dried
51.66
10.5
22.42
97.86
1359.
5
Parsley
Parsley,raw
6.33
0.16
2.97
6.2
1640
Coriander leaf
Coriander
leaf,dried
52.1
6.355
21.93
42.46
1359.
5
4037
7. Shah, Jaimeel M., and Lokesh Sahu. "A Hybrid Based Recommendation
System based on Clustering and Association [J]." Binary Journal of Data
Mining & Networking 5.1 (2015): 36-40.
8. Subramaniyaswamy, V., and R. Logesh. "Adaptive KNN based recommender
system through mining of user preferences." Wireless Personal
Communications 97.2 (2017): 2229-2247.
9. Darvishy, Asghar, et al. "A Customized Non-Exclusive Clustering Algorithm
for News Recommendation Systems." Journal of University of Babylon for
Pure and Applied Sciences 27.1 (2019): 368-379.
10. Ahmed, Muyeed, Mir Tahsin Imtiaz, and Raiyan Khan. "Movie
recommendation system using clustering and pattern recognition
network." 2018 IEEE 8th Annual Computing and Communication Workshop
and Conference (CCWC). IEEE, 2018.
11. Phanich, Maiyaporn, Phathrajarin Pholkul, and Suphakant Phimoltares.
"Food recommendation system using clustering analysis for diabetic
patients." 2010 International Conference on Information Science and
Applications. IEEE, 2010.
12. Gaikwad, D. S., et al. "Food Recommendation System." International
Research Journal of Engineering and Technology 4.01 (2017).
13. Elahi, Mehdi, et al. "Interaction design in a mobile food recommender
system." CEUR Workshop Proceedings. CEUR-WS, 2015.
14. Sawant, Sumedh, and Gina Pai. "Yelp food recommendation system." (2013).
15. Maia, Rui, and Joao C. Ferreira. "Context-aware food recommendation
system." Context-aware food recommendation system (2018): 349-356.
16. Oyelade, O. J., O. O. Oladipupo, and I. C. Obagbuwa. "Application of k Means
Clustering algorithm for prediction of Students Academic Performance." arXiv
preprint arXiv:1002.2425 (2010).
17. Basim Amer Jaafar, Methaq Talib Gaata, Mahdi Nsaif Jasim. "Home
appliances recommendation system based on weather information using
combined modified k-means and elbow algorithms." Indonesian Journal of
Electrical Engineering and Computer Science (2020): Vol. 19, No. 3,
September 2020, pp. 1635~1642ISSN: 2502-4752, DOI:
10.11591/ijeecs.v19.i3.pp1635-1642.
18. Kodinariya, Trupti M., and Prashant R. Makwana. "Review on determining
number of Cluster in K-Means Clustering." International Journal 1.6 (2013):
90-95.
19. Mohammed Ibrahim and Mahdi Nsaif Jasim. "New Modified Dynamic
Clustering Algorithm," Journal of Engineering and Applied Sciences, vol. 14,
no. 18, pp. 6742-6746, 2019.
20. Danganan, Alvincent E., Ariel M. Sison, and Ruji P. Medina. "OCA:
overlapping clustering application unsupervised approach for data analysis."
Indonesian Journal of Electrical Engineering and Computer Science 14.3
(2019): 1471-1478.
21. Oyelade, O. J., O. O. Oladipupo, and I. C. Obagbuwa. "Application of k Means
Clustering algorithm for prediction of Students Academic Performance." arXiv
preprint arXiv:1002.2425 (2010).
22. Lailiyah, Siti, Ekawati Yulsilviana, and Reza Andrea. "Clustering analysis of
learning style on anggana high school student." Telkomnika 17.3 (2019):
1409-1416.
4038
23. Mahdi, Muhammed U. "Determining Number & Initial Seeds of K-Means
Clustring Using GA." Journal of Babylon University and Applied Sciences
18.3 (2010).
24. Arellano-Verdejo, Javier, et al. "Efficiently finding the optimum number of
clusters in a dataset with a new hybrid cellular evolutionary algorithm,"
Computación y Sistemas, vol. 18, no. 2, pp. 313-327, 2014.
25. Yuan, Chunhui, and Haitao Yang. "Research on K-value selection method of
K-means clustering algorithm." J—Multidisciplinary Scientific Journal 2.2
(2019): 226-235.
26. Subbalakshmi, Chatti, Rishi Sayal, and H. S. Saini. "Cluster Validity Using
Modified Fuzzy Silhouette Index on Large Dynamic Data Set." Computational
Intelligence in Data Mining. Springer, Singapore, 2020. 1-14.
27. Subbalakshmi, Chatti, et al. "A method to find optimum number of clusters
based on fuzzy silhouette on dynamic data set." Procedia Computer Science
46 (2015): 346-353.
28. Florian Pecune ,Lucile Callebert , and Stacy Marsella, Recommender System
for Healthy and Personalized Recipe Recommendations, HealthRecSys’20,
September 26, 2020, Online, Worldwide.
29. Arnawa, I.K., Sapanca, P.L.Y., Martini, L.K.B., Udayana, I.G.B., Suryasa, W.
(2019). Food security program towards community food consumption.
Journal of Advanced Research in Dynamical and Control Systems, 11(2), 1198-
1210.
30. Normatova, S. A. ., Botirov, M. T. ., Ruzmatova, K. K. ., & Mamarasulov, J. O.
ugli . (2021). Hygienic basis for contamination of food products and
production of dairy products until 2030. International Journal of Health &
Medical Sciences, 4(1), 123-128. https://doi.org/10.31295/ijhms.v4n1.1592
