Grenze International Journal of Engineering and Technology, Jan Issue

Cattle Disease Prediction using Machine Learning: A

Systematic Review
Deeksha G Mesta1, Janani R2, Nydile M N3 and Prof. Naveen Kumar M R4
1-4
Information Science and Engineering, Vidyavardhaka College of Engineering, Mysuru, India
Email: [email protected], [email protected], [email protected], [email protected]

Abstract—The potential benefits of significant livestock health diagnosis is enormous. Animal

husbandry being the most important aspect of India’s agricultural growth, farmers suffer on a
regular basis as a result of their cattle's bad health and lack of qualified veterinary doctors in
their region. At a breakneck pace of Big data and Machine Learning, data mining and analysis
are implemented widely in the field of animal husbandry because of the fact that disease and
bioterrorism constitute a danger to cattle produce and food supply’s safety. In this review
paper, we will be comparing the different data science techniques.

Keywords—Data Mining, Machine Learning, Cattle Disease.

I. INTRODUCTION
In view of rising demand driven by rising revenue increased health consciousness, and a growing population, the
slowing in the dairy business is highly concerning. According to the National Dairy Development Board
(NDDB), milk demand would exceed 180 million tonnes by 2022. To meet demand over the next 15 years, an
average exponential increase of 5 million tonnes per year is required, which is more than double the average
incremental rate achieved during the previous century[1]. If there is a lack of local milk production, India will
have to rely on imports from other countries for its dairy needs. As a result, focusing on local dairy expansion is
critical.
Identification of symptoms, cattle diseases, and providing proper treatments is becoming a tedious job in the
current medical field. In real-time, it is difficult to handle the cattle disease symptoms and disease types as
animals can't explain the problems or pain that they are facing. The manual process of identifying the cattle
disease and treatment is too complex and time-consuming and also expensive[2]. The existing system is a
manual process where doctors diagnose animals, identify diseases, and give treatment. In foreign countries, they
use some advanced systems such as the MYCIN expert system, IBM Watson, etc. The systems which use
advanced technologies gather the data, retain it in the database and restore it back whenever required. This
makes the medical practitioner job easy to treat cattle diseases in an advanced and effective way.
To our knowledge, no work has been done that focuses on giving a thorough classification of Machine Learning
techniques in the prediction of cattle disease. Furthermore, none of the research has thoroughly compared and
analyzed machine learning algorithms in the prediction of cattle disease based on the benefits and drawbacks of
ML-based techniques
This review paper organised as section(i) Related Works, section(ii) Background, section(iii) Comparative
Studies, and section(iv) Conclusion.

Grenze ID: 01.GIJET.9.1.613

© Grenze Scientific Society, 2023
 Figure. 1 Systematic Review Process

II. RELATED WORKS

Using Data Mining Techniques and Cattle Health Monitoring System many research papers have been published
on disease prediction. Different data mining methods are used for prediction and obtaining dissimilar outcomes
for different models.
In their report "Developing Mobile Intelligent System For Cattle Disease Diagnosis and First Aid Action
Suggestion," Wiwik Anggraeni, A.Muklason, A.F.Ashari, Wahyu A., and Darminto describe the effort of
designing a system, i.e. mobile intelligence, for diagnosing illnesses in cattle and providing first-aid suggestions,
which was done using a fuzzy neural network. Another system based on the Fuzzy Neural System was
created[1]. In their paper "A Fuzzy Expert System to Diagnose Diseases with Neurological Signs in Domestic
Animals," Mahdi Jampour, Mohsen Jampour, Maryam Ashourzadeh, and Mahdi Yaghoobi used the Fuzzy
Expert System, which can detect and assess the absence of involvement of any disease with neurological
symptoms[2].
Kunja Bihari Swain, Satyasopan Mahato, Meerina patro, Sudeepta Kumar pattnayak in their work on “Cattle
health monitoring system using Arduino and LabVIEW for early detection of diseases” they used Arduino UNO,
Arduino NANO, Xbee module, and Sensors of various types are used to measure the bodily parameters of
cattle[3].
Francisco Gomez, Jeisson Prieto, Juan Galvis, Fausto Moreno, Jimmy Vargas in their work on “Identification of
Super-Spreaders of Foot-and-Mouth Disease in the cattle transportation network: The 2018 outbreak case in
Cesar (Colombia)” by using cattle transportation network. The centrality of the network may provide
information on FMD epidemiology as well as the dynamics of animal movement[4].
"Identification of Acidosis Disease in Cattle Using IoT," by Fatih and Kamil. pH sensor and waterproof
temperature sensors were employed with the Internet of Things (IoT) technology. By monitoring and
documenting animal diet and behavior on a regular basis, this gadget can detect sickness. Another IoT-based
solution has been developed[5]. In their paper "Cattle disease diagnosis using Prediction Technique" Noone
Vijay Kishan, Sai Trinath Y, Sandeep Kavalur, Sangamesha V, and Mr. Sumanth Reddy employed IoT and Data
Mining. Sensor technologies were used to chart the basic aspects of animal activity, such as heartbeat[6]. The
data was aggregated and transmitted to a data mining algorithm to validate the disease. Another IoT-based
solution has been developed. Emi Mwaka Katemboh, Raed Abdulla, Vikneswary Jayapal, Sathish Kumar
Selvaperumal, Dhakshyani Ratnadurai in their work on “Integrated Animal Health Care Using IoT'' used the
sensor and the microcontroller that is the Arduino that has been mounted on a LoRa RFM shield and this acts as
a node where data is sent to which can predict the disease[7].
Varun Garg, Kritika Garg in their work on “Early-Stage Disease Detection Platform in Cattles” used Neural
Network which is used to provide economic medical solutions[8].

831
Lauren Wisnieski, Steven Pierce, Bo Norby, Tyler Becker, Jeff Gandy, and Lorraine Sordillo employed Support
Vector Machine in their project "Cattle Disease Auxiliary Diagnosis and Treatment System Based on Data
Analysis and Mining." They used electronic data and data analysis technology to construct an intelligent cow
sickness detection system[9]. Another solution to the problem was established by Daksh Ashar, Amit Kanojia,
and Rahul Parihar in their work on "Livestock Sickness Prediction System," which employed Support Vector
Classifier to predict the data and inform the livestock owner if the predicted disease may cause sudden death[10].
In this paper S. Shahinfar, D. Page, J. Guenther, V. Cabrera, P. Fricke, and K. Weigel, “Prediction of
insemination outcomes in Holstein dairy cattle using alternative machine learning algorithms,” have addressed
reproduction in cattle by developing tool to make decisions on cattle breeding. The Random forest algorithm has
given better results than naïve Bayes, Bayesian network and decision tree algorithms. The parameters which they
have considered are gain ratio (GR) Herd-year-month (HYM) and Information gain (IG). They have achieved an
accuracy of 75.6 and 73.6 in primiparous and multiparous respectively[11].
In this paper A. D. Sunny, S. Kulshreshtha, S. Singh, M. Ba, and H. Sarojadevi, “Disease Diagnosis System By
Exploring Machine Learning Algorithms,” have developed a system which diagnose the cattle disease based on
the input text data. The authors have compared different machine learning algorithms like naïve bayes, k-nearest
neighbor and aprior algorithms where Aprior algorithm has yielded good result and naïve bayes are also good
depending upon the dataset which they have considered[12].
In this paper L. Wan, W. Bao, A. Disease, D. Expert, S. Based, and T. Ifip, “Animal Disease Diagnoses Expert
System have proved the practicality of Support vector machine (SVM) which can be used to livestock breeding
of cattle and developed the expert system that can diagnose the diseases. The experiment carried out on a small
dataset which yielded the good result which is the limitation of the work[13].
In this article D. Wu et al., “Lameness detection of dairy cows based on the YOLOv3 deep learning algorithm
and a relative step size characteristic vector,” have proposed yolov3 deep learning algorithm and relative step
size characteristics to identify the lameness detection in cows. Lameness detection was done using the videos
where LSTM, SVM, KNN and DTC are used to classify the lame and non-lame legs of the cow. The LSTM
outperformed all other algorithms with 98.57 accuracy. The yolov3 model is used to identify the target position
of the legs in videos. The authors have proposed the extension of the work can be used for automatic detection of
lameness detection in cows. Lameness can increase the risk of premature elimination of dairy cows, and can
cause huge economic losses. Data on the acceleration when cows walked were collected by acceleration sensors
fixed to the hind legs, and gait analysis was carried out[14]. In order to realise the detection and classification of
lame cows and nonlame cows, a method based on the YOLOv3 deep learning algorithm and a relative step size
characteristic vector is proposed in this study. Intelligent lameness detection is important for improving cow
welfare. A method based on YOLOv3 deep learning algorithm and relative step size characteristic vector is
proposed to classify lame and non-lame cows. Relative step sizes of cow's front and rear legs were calculated
based on leg coordinates, and the relative step size characteristic vector was constructed. Finally, a trained Long
Short-Term Memory (LSTM) classification model was used to classify lame and non-lame cows based on the
characteristic vector.
In this paper, H. Sadeghi, H. S. Braun, B. Panti, G. Opsomer, and O. B. Pascottini, “Validation of a deep
learning-based image analysis system to diagnose subclinical endometritis in dairy cows,” have assessed the
polymorph nuclear leukocyte (PMN) proportion of endometrial samples using deep learning. When compared to
naphthol and Diff-Quick the deep learning algorithms have yielded good results. In endometrial cytology, the
deep learning algorithms are reliable and deduce the bias in subclinical endometriosis[15].
In this article, V. Sturm, D. Efrosinin, M. Öhlschuster, E. Gusterer, M. Drillich, and M. Iwersen, “Combination
of sensor data and health monitoring for early detection of subclinical Ketosis in dairy cows,” have proposed an
algorithm that can predict the occurrence of subclinical ketosis in cows. The flexible algorithm uses TSC (TIME
SERIES CLASSIFIER) with a traditional classification algorithm to classify the data from the sensors. The
results show that the method used for automatic detection of diseases in dairy cows performed good. Future
elaborated machine learning algorithms can be used to early detect subclinical ketosis[16].
In this paper, N. Misk, T. Misk, A. El-Khamary, and M. Semeika, “A retrospective study of surgical affections
of mammary glands in cattle and buffaloes and their management in the field,” have used non deep learning
algorithm like random forests and support vector machines to classify the data. The experiment was conducted
on image data where random forest achieved the accuracy of 95.20%. Since data is the challenge to train the
deep models and it consumes more computation the authors have adopted non- deep models for both classifying
and training[17].
In this article, A. Hidalgo, F. Zouari, H. Knijn, and S. van der Beek, “Prediction of postpartum diseases of dairy
cattle using machine learning,” have used machine learning algorithm to predict occurrence of postpartum

832
diseases by using prepartum data. In this study they have used random forest algorithm to predict the postpartum
diseases. The data set were real time data taken from Dutch farms from 2009 to 2016. The algorithm
outperformed other state of art algorithms in machine learning[18].
In this article, “Algorithms for Detecting Cattle Diseases at Early Stages and for Making Diagnoses and Related
Recommendations” have developed an algorithm to identify infectious, parasitic, and nervous diseases in large
horned cattle. They have taken a data from 5farms. They wrote the code in a software package “Matlab
R2019b”[19].
In this article, Y. Ali, A. Farooq, T. M. Alam, M. S. Farooq, M. J. Awan, and T. I. Baig, “Detection of
Schistosomiasis Factors Using Association Rule Mining” have developed detection technique by using Apriori
Algorithm to detect the Schistosomiasis disease in an early stage. They collected dataset from the record of 4136
(2674 male and 1462 female) patients, diagnosis with advance disease in them[20] and another article based on
Apriori algorithm, M. Shamsul Arifin, M. Golam Kibria, A. Firoze, M. Ashraful Amini, and H. Yan, “Detection
and Prediction of Frequent Diseases in India through Association Technique using Apriori Algorithm and
Random Forest Regression” has developed a technique to identify the frequency of the data[21] and there is
another article which is based on clustering method, “Prediction of Heart Disease by Clustering and
Classification Techniques Prediction of Heart Disease by Clustering and Classification Techniques” have
developed an system from which the heart disease can be predicted[22].
In this article, A. A. Elngar, R. Kumar, A. Hayat, and P. Churi, “Intelligent System for Skin Disease Prediction
using Machine Learning” have developed a system for detecting the skin disease and provide treatment for the
same. They have used CNN and SVM algorithm for detecting the skin disorders[23] and there is another
prediction system article, M. Shamsul Arifin, M. Golam Kibria, A. Firoze, M. Ashraful Amini, and H. Yan,
“Dermatological disease diagnosis using color-skin images” where they used backpropogation neural network to
detect the skin diseases in an early stage[24].
In this article, N. Wagner et al, “Machine learning to detect behavioural anomalies in dairy cows under subacute
ruminal acidosis” to detect abnormal behaviour in dairy cows suffering from subacute ruminal acidosis. They
used K Nearest Neighbours for Regression (KNNR); Decision Tree for Regression (DTR); MultiLayer
Perceptron (MLP); Long Short-Term Memory (LSTM) algorithms[25].
In this article, W. Xu et al, “Prediction of metabolic status of dairy cows in early lactation with on-farm cow data
and machine learning algorithms” here based on plasma metabolites and metabolic hormones, dairy cows might
be grouped according to their metabolic condition. They used Decision Tree, Naive Bayes, Bayesian Network,
Support Vector Machine, ANN, Bootstrap Aggregation, Random Forest, and K-Nearest Neighbor[26].
“Identifying Health Status in Grazing Dairy Cows from Milk Mid-Infrared Spectroscopy by Using Machine
Learning Methods”, B. C. Hernández, N. Lopez-Villalobos, and M. Vignes by have developed a technique in
which they can identify the health status of cows. They have used Random forest, support vector machine, neural
network, convolutional neural network[27].
“Assessing machine learning techniques in forecasting lumpy skin disease occurrence based on meteorological
and geospatial features” E. Afshari Safavi, using ANN may be used to accurately predict the occurrence of
LSDV infection using geographical and meteorological data[28].
“MasPA: A Machine Learning Application to Predict Risk of Mastitis in Cattle from AMS Sensor Data” N.
Abdul Ghafoor and B. Sitkowska, used random forest algorithm can forecast the likelihood of mastitis in cattle
based on their udder's inhale and exhale limitations and temperature[29].
In article, U I Bajwa*, S Alam, N ul Haq, N I Ratyal and M Waqas Anwar, V, “Skin Disease Classification using
Neural Network” main goal of their research is to create an automated method for skin disease categorization
that can assist a general practitioner in detecting and classifying lesions. They used ANN for classification[30].
In machine learning, a dataset is essentially a collection of data bits that can be analyzed and predicted by a
computer as a single entity. This implies that the data gathered should be homogeneous and understandable to a
machine that does not see data in the same manner that people do. After gathering the data, it's crucial to clean
and complete it, as well as label it with significant computer-readable tags. Additionally, a useful dataset should
meet specific quality and quantity requirements. The dataset must be relevant and well-balanced for a smooth
and quick training experience.

III. BACKGROUND
A) MACHINE LEARNING
Machine learning is classified as a multidisciplinary branch of computer science. The main goal of machine
learning is to find undiscovered facts, hidden insights, correlations, patterns, and knowledge in massive data sets

833
that are difficult to find and identify using traditional statistical approaches; the design and development of data-
learning systems. The main steps involved in machine learning are preprocessing, modeling, and post-processing
of data.
Machine learning is classified into three categories:
Supervised Machine learning:-The algorithms that use externally given cases to build a broad hypothesis and
make use of it in the future for predictions are called supervised machine learning algorithms. One of the
distinguishing features is their capacity to use annotated data. SML algorithms use class labels for the
classification process. There is a wide range of supervised algorithms some of them are: Linear Classifiers,
Support Vector Machine Logistic Regression, Naïve Bayes Classifier, Perceptron, Quadratic Classifiers,
Decision Tree, K-Means Clustering, Boosting, Random Forest (RF); Bayesian Networks, Neural Networks
and so on

Figure. 2 Supervised machine learning model

Support Vector Machine: SVM being a supervised machine learning algorithm was formulated by Vapnik
et al. in 1995. Support Vector Machine (SVM) models are similar to traditional multilayer perceptron neural
networks. The concept of “margin”(separates two different classes ) is used in SVM. The margin is maximized
to produce the biggest possible space between the separating hyperplane and instances and thus reduces the
predicted generalization error. SVM first calculates the low dimensional space. Then the kernel function is used
to map high dimensional feature space and input space. Finally, in the high dimensional feature space, the best-
classified hyperplane is built[9].
Input: Consider the training dataset T with multiple symptoms, where symptoms of diseases and category of
diseases are denoted by xi and yi respectively. f(x) is the decision function which produces the optimal
classification plane is shown below
Neural Networks: Neural Networks (NN) can execute many classifications and/or regression tasks at the same
time, despite the fact that most networks can only do one. In the great majority of situations, there will be a
single output variable for the network. However, with respect to many state classification problems, it equates to
the number of output units[8].

Figure. 3 Neural Network Equation

Fuzzy Neural Networks: A simple fuzzy neural network produces a Gaussian function by narrowing down the set
of input vectors, which is then correlated to the target label. Secondly, in order to obtain a fuzzy truth, the
previously found gaussian function is used, hence classifying the unknown feature vector. Thirdly, to find the
class of input feature vectors, fuzzy truths are maximized [3].

834
 Figure. 4 Simple Fuzzy Neural Network

Naive Bayes Classifier:Naive Bayes Classifier is a classification method based on Bayes' Theorem and the
condition of predictor independence. A Naive Bayes classifier, in simple terms, asserts that the existence of one
feature in a class is independent of the presence of any other feature. The Naive Bayes model is simple to
construct and is exceptionally effective for huge datasets. Naive Bayes is recognized to outperform even the
most advanced classification systems due to its simplicity.

Figure. 5 Naive Bayes Equation

Random Forest Classifier:A random forest is a meta estimator that uses averaging to increase predicted accuracy
and control over-fitting by fitting a number of decision tree classifiers on various sub-samples of the dataset. If
bootstrap=True (default), the sub-sample size is regulated by the max samples argument; otherwise, the entire
dataset is utilized to create each tree.

Figure. 6 Random Forest Classifier

835
Decision Tree Classifier: Decision Tree is a supervised learning technique that may be used to solve both
classification and regression problems, however, it is most commonly used to solve classification issues. Internal
nodes represent dataset attributes, branches represent decision rules, and each leaf node provides the result in this
tree-structured classifier. The Decision Node and the Leaf Node are the two nodes of a Decision tree. Leaf nodes
are the output of those decisions and do not contain any more branches, whereas Decision nodes are used to
make any decision and have several branches. The decisions or tests are made based on the characteristics of the
given dataset.

Figure. 7 Decision Tree Classifier

B) Unsupervised Learning
This Learning technique will not make use of training datasets to build models, but instead, find the hidden
patterns in the given dataset. In this technique, class labels are not present. Unsupervised techniques cannot be
applied to classification or regression tasks. It covers the underlying patterns based on similarities to display the
dataset in a compact fashion. Different algorithms that come under unsupervised methodologies are “eclat
algorithm”, “FP Growth algorithm”, “AIT algorithm” and many more.

Figure. 8 Unsupervised learning model

C) Reinforcement Learning
Machine Learning includes the field of reinforcement learning. It's all about taking the right steps to maximize
your reward in a given situation. It is used by a variety of software and computers to determine the best feasible
action or path in a given situation. Reinforcement learning differs from supervised learning in that supervised
learning includes the answer key, allowing the model to be trained with the correct answer, whereas
reinforcement learning does not include an answer and instead relies on the reinforcement agent to decide what
to do to complete the task. It is obligated to learn from its experience in the absence of a training dataset.

836
 Figure. 9 Reinforcement Learning

IV. COMPARATIVE STUDY

Author Year Methodology Dataset Description Limitation

M. Jampour, 2011 Fuzzy Expert 20 test board of The fuzzy logic system method is Used only to Diagnose
M. Jampour, System animal used to determine and compute Diseases with
M. the absence of participation of Neurological Signs
Ashourzadeh, each probable disease with
and M. neurological symptoms
Yaghoobi

M. Shamsul 2012 Backpropagatio Image dataset The system uses colour image This system gave best
Arifin, M. n Neural processing techniques, k-means results for image dataset
Golam Network clustering, and colour gradient only.
Kibria, A. approaches to identify the sick
Firoze, M. skin in terms of machine
Ashraful intervention. The system uses
Amini, and feedforward backpropagation
H. Yan artificial neural networks to
classify diseases.

W. 2013 Fuzzy Neural 250 medical The purpose of this study is to Android app is developed,
Anggraeni, Network records for three demonstrate the work that has visualization problem on
A. Muklason, different cattle gone into establishing a mobile training data-sets.
A. F. Ashari, diseases i.e. intelligent system for diagnosing
A. Wahyu, Mastitis, foot and cattle ailments and suggesting
and Darminto mouth disease, first-aid actions
and Septicemia
Epizootic.

S. Shahinfar, 2014 Random forest 26 Wisconsin It is very difficult to predict the The outcome for
D. Page, J. algorithm dairy farms that fertilization results of individual lactating dairy
Guenther, V. were enrolled in lactating cows. we can use cows is extremely
Cabrera, P. the Alta information about health, breeding difficult, information
Fricke, and Genetics. history, production level, and regarding health,
K. Weigel other environmental reproductive history,
characteristics to identify a very production level, and
fertile subgroup of other environmental
heads. features can be used to
identify highly fertile
subsets of cows

L. Wan, W. 2014 Support Vector Supported vector machines The difficulty of the rapid
Bao, A. Machine (SVMs) can be used for breeding diagnosis processed
Disease, D. and developing expert systems because of the complexity

837
Expert, S. that can diagnose diseases. and diversity of the
Based, and T. Testing conducted on a small data symptoms of animal
Ifip set for precise results is the limit diseases.
of the work.

V. Garg, K. 2016 Neural Network Differing types of Machine learning was applied to Achieved accuracy is
Garg skin diseases detect cow ailments, resulting in a more only for skin related
from color Image cost-effective medical treatment diseases.

K. B. Swain, 2017 Arduino UNO, Data from Designed a system that allows Only used to monitor the
S. Mahato, Arduino sensors farmers to monitor and compare cattle health condition.
M. Patro, S. NANO, Xbee the current health parameters of
K. Pattnayak module their livestock to the conventional
healthy reference parameters

F. Gomez, J. 2018 Cattle The cow transportation network is Only used to monitor the
Prieto, J. Transportation computed using several centrality “foot-and-mouth disease”.
Galvis, F. Network measures, each of which may Small data-sets used
Moreno, J. (ranking nodes) provide useful information about
Vargas the animal movement

A. D. Sunny, 2018 Naive Bayes Disease- The authors compared different The disease being
S. and Apriori Symptom machine learning algorithms like predicted based on the
Kulshreshtha, Knowledge data naive algorithm, neighbor input symptoms, using
S. Singh, M. set prepared by algorithms, and aprior where the multiple algorithms.
Ba, and H. New York Aprior algorithm performed well
Sarojadevi Presbyterian and the naive algorithms were also
Hospital of good based on the data set they
patients admitted studied. consider.
during 2004

A. Hidalgo, 2018 Random forest The data set were Predict the occurrence of We envisage that big data
F. Zouari, H. real time data postpartum diseases using is out there to stay and it
Knijn, and S. taken from Dutch prenatal data. In this study, they has different applications.
van der Beek farms from 2009 used a random forest algorithm to
to 2016. predict postpartum morbidity.
This algorithm has performed
better than other modern
algorithms in machine learning.

F. Basciftci, 2019 Internet of Data from Regular monitoring and Only used to monitor one
K. A. Things (IoT) pH sensors documenting of animal nutrition disease “Acidosis
Gunduz sensor, and daily behavior can help to Disease”. Sensors used
Waterproof identify the condition for monitoring leads to
temperature less accurate results.
sensors

Y. Ali, A. 2019 Apriori Data from 4136 The major goal of this study is to This system disease
Farooq, T. Algorithm patients were investigate the disease in depth, detection was only for the
M. Alam, M. diagnosed with assess connections between human beings they didn’t
S. Farooq, M. advanced disease factors, and analyse the concentrate on animals.
J. Awan, and schistosomiasis disease using association rules in
T. I. Baig illness (2674 schistosomiasis patients.
males and 1462
females).

Singh, Reetu 2019 K-Means clinical dataset Goal was to predict the heart The accuracy of
Rajesh, E. Clustering diseases by clustering and clustering and
classifying the data. classification is reduced
when some data remain
un-clustered or wrongly
clustered.

W. Xu et al. 2019 Random Forest collected from 2 By using machine learning Accuracy was low.

838
 and SVM earlier studies algorithm, Predicting of metabolic
with proto- status of dairy cows in early
col number lactation with on-farm cow data.
2010026 (study
1; van Knegsel et
al., 2014;
Chen et al., 2016)
and 2014125
(study 2; van
Hoeij et al.,
2017). The

L. Niu, C. 2020 Support Vector Multi-source cow To create the intelligent diagnosis Large dataset is required
Yang, Y. Du, Machine (SVM) electronic system for cattle diseases, a huge to give accurate results.
L. Qin, B. Li medical record quantity of multi-source cow Only when the symptoms
data electronic medical record data was are less accuracy is more.
collected and data analysis and
mining technologies was applied

D. Wu et al. 2020 Support Vector Walking crawl The lameness detection was The algorithm proposed
Machine acceleration data performed using videos in which in this study could
is collected by LSTM, SVM, KNN and DTC effectively overcome
accelerationSens were used to classify lame and different forms of
ors were attached non-crippled cows. The yolov3 interference such as
to the hind legs model is used to determine the occlusion, light change
target position of the foot in the and near-view objects,
video. They propose that an and can achieve rapid and
extension of the work could be accurate detection of leg
used to automatically detect targets.
lameness in cows.

V. Sturm, D. 2020 Naive Bayes Dataset approved This may predict the occurrence Early detection of SCK is
Efrosinin, M. from University of subclinical ketosis in cows. The not possible.
Öhlschuster, of Veterinary flexible algorithm uses TSC with
E. Gusterer, Medicine Vienna, traditional classification algorithm
M. Drillich, Austria and to classify sensor data. The results
and M. Slovakian show that the automatic disease
Iwersen Regional detection method on dairy cows
Veterinary Food gives good results. In the future,
Administration. developed machine learning
algorithms can be used for early
detection of subclinical ketosis.

M. Bhanu 2020 Apriori National Health Analysing the data to predict the Dataset they have taken
Sridhar Algorithm and Profile by the diseases. was very less.
P. Aiswarya Random Forest Central Bureau of
L. Kavitha Regression Health
Intelligence,
Government of
India, along with
the collaboration
of World Health
Organization
(WHO)

N. Wagner et 2020 K Nearest Clinical Dataset Using algorithms they have Less accuracy
al. Neighbours for analyzed the disease behavior
Regression, characteristics.
Decision Tree
for Regression,
MultiLayer
Perceptron,
Long Short-
Term Memory

UI 2020 ANN 588 images Goal of this automated system is They didn’t analyze of
Bajwa*, S containing 6907 to classify the disease in skin. different skin tones on the

839
Alam, N u lesion regions performance of lesions
Haq, N I detection and
Ratyal and M classification system
W Anwar

N. V. Kishan, 2021 Random forest 95 symptoms of The basic characteristics of animal Doesn't predicts the
S. T. Y, S. Classifier, cow activity, such as heartbeat, were relationship between
Kavalur Decision Tree charted using sensor technologies. symptoms, disease and
Classifier, To validate the disease, the data treatments.
Naïve Bayes was compiled and fed into a data
Classifier mining model

D. Ashar, A. 2021 Support Vector Animal Dataset A multi-class classification Accuracy achieved was
Kanojia, R. Classifier algorithm was utilised to provide less.Time consuming as
Parihar, S. preventative actions based on the the prediction is made for
Kudoo disease forecast, as well as to goat, sheep and cow using
inform the livestock owner if the a single model.
disease indicated could result in a
sudden death

D. Y. Pavkin, 2021 5 large farms: They looked at the relationships They concentrated onyl
A. S. “Ecocorm” Ltd. between the temperature and pH on the cattles not on the
Dorokhov, F. (Voronezh of the cows' rumens, their other animals.
E. region, Russia), locomotor activity, and
Vladimirov, “Research Center environmental factors including
I. M. of Traditional air temperature and relative
Dovlatov, Technologies” humidity in the cowsheds. After
and K. S. Ltd. (Ivanovo that, they created graphs of
Lyalin region, Russia), functionalities that are intertwined
Agricultural for the first time, algorithms were
enterprise compiled as a result of the
“Novomarkovsko research. Detecting viral, parasitic,
ye” Ltd. and neurological illnesses was
(Voronezh possible.
region, Russia),
Agricultural
enterprise
“Voshchazhnikov
o” Ltd.
(Yaroslavl
region, Russia)

A. A. Elngar, 2021 SVM and CNN 3000 images To detect the skin disease at early This was helpful for only
R. Kumar, A. which collected stage they have developed a the image data.
Hayat, and P. from a lot of prediction system from which
Churi sources like they will be able to get the
Beni-Suef treatment for the particular skin
University disease.
Hospital, Cairo
University
Hospital

M. Shamsul 2021 Apriori National Health a strategy for extracting This work can be
Arifin, M. algorithm and Profile by the interesting patterns from health expanded to include
Golam Random Forest Central Bureau of data using the Association pandemics that are
Kibria, A. Classfier Health Method. unforeseen and
Firoze, M. Intelligence, predictions that can help a
Ashraful Government of country prepare for the
Amini, and India, along with invisible adversary.
H. Yan the collaboration
of World Health
Organization
(WHO), various
Central
Ministries and all

840
 the state/union
territory health
departments

B. C. 2021 random forest, 1909 milk Identifying the health status of the Less number of dataset
Hernández, support vector samples cows.
N. Lopez- machine, neural
Villalobos, network (NN),
and M. convolutional
Vignes neural network
and

N. Abdul 2021 Random Forest 6600 entries To predict the risk of mastitis in Less accuracy
Ghafoor and (three entries per cattle
B. Sitkowska cattle) for cattle
with 15 attributes

H. Sadeghi, 2022 Oculyze MUH A large dataset of This system evaluates the To improve the
H. S. Braun, system cells of the polymorphonuclear leukocyte performance and
B. Panti, G. different classes (PMN) percentage of endometrial robustness of the Oculyze
Opsomer, has been used to samples using deep learning. MUH, an expansion of
and O. B. obtain the Compared to naphthol and the dataset to train the
Pascottini retrained model DiffQuick, the deep learning system with an extensive
algorithms perform well. set of images is
In endometrial cytology, deep warranted.
learning algorithms are reliable
and inferentially biased in
subclinical endometriosis.

N. Misk, T. 2022 Random Forest Image data The experiment was conducted on The results of this study
Misk, A. El- and Support image data. Since the data is demonstrate that
Khamary, Vector Machine challenging to train deep models treatment of most teat and
and M. and it is more computationally udder affections is
Semeika intensive, the authors used non- feasible in field settings
deep models for classification and and veterinarians can
training. expect satisfactory results
after interventions.

E. Afshari 2022 ANN LSDV infection Using geospatial and climatic Difficult to predict for
Safavi data parameters, ANN can be used to larger dataset.
predict the occurrence of LSDV
infection with great precision. In
locations where the occurrence of
LSDV infection is a high risk,
using the predicting capacity of
these methodologies could be very
useful in implementing screening
and awareness initiatives, as well
as taking preventive measures like
vaccination.
Keywords

V. PROPOSED METHOD
Disease remains the major cause of death. Several cattle have died as a result of an illness for which the proper
treatment is unknown. So we did a comparison research and discovered that in many research papers they
have used a different algorithm such as Aprior. As a result, we'll use a number of unsupervised algorithms to
provide accurate results. The proposed system takes a symptom as input and predicts the disease as well as
therapy options. We used a few dataset to try to predict illness. The eclat algorithm was found to be 0.38
millisecond efficient and 86.2%.

VI. PRILIMINARY R ESULT

Cattle disease can be diagnosed using machine learning algorithms, which aids veterinary doctors in providing
therapy. We must use the algorithm on the training dataset. As we discovered throughout our comparison

841
research, majority of them used a supervised technique to predict sickness. As a result, we will attempt to build
the project by applying an unsupervised learning technique

Figure. 10 Dataset Classification Flowchart

VII. CONCLUSION

Several cattle diseases result in a decrease in milk yield. Many researches were conducted to predict these
diseases with the help of machine learning models. But the performance of the technique depends upon the
missing and noisy data in the dataset. Accuracy and confusion matrix relies upon the above-mentioned factors
which can be avoided by using accurate statistical approaches to clean the dataset. We have reviewed the
drawbacks and advantages of each algorithm in the comparative study section. According to the survey Support
Vector Machines produced more accurate results and were found useful in many areas of animal husbandry
when compared to neural networks and IOT. The performance of algorithms vary greatly between different
datasets because of their data source, techniques used to preprocess the data and selection the parameters. But
they can use any algorithm which is suitable for their dataset. However, there is still scope in the veterinary field
for experts to work. We will be implementing a project where we will be comparing our algorithm with Aprior
algorithm.

REFERENCES
[1] W. Anggraeni, A. Muklason, A. F. Ashari, A. Wahyu, and Darminto, “Developing mobile intelligent system for cattle
disease diagnosis and first aid action suggestion,” Proc. - 2013 7th Int. Conf. Complex, Intelligent, Softw. Intensive
Syst. CISIS 2013, pp. 117–121, 2013, doi: 10.1109/CISIS.2013.27.
[2] M. Jampour, M. Jampour, M. Ashourzadeh, and M. Yaghoobi, “A fuzzy expert system to diagnose diseases with
neurological signs in domestic animal,” Proc. - 2011 8th Int. Conf. Inf. Technol. New Gener. ITNG 2011, pp. 1021–
1024, 2011, doi: 10.1109/ITNG.2011.217.
[3] K. B. Swain, S. Mahato, M. Patro, and S. K. Pattnayak, “Cattle health monitoring system using Arduino and LabVIEW
for early detection of diseases,” Proc. 2017 3rd IEEE Int. Conf. Sensing, Signal Process. Secur. ICSSS 2017, no. May,
pp. 79–82, 2017, doi: 10.1109/SSPS.2017.8071569.
[4] F. Gomez, J. Prieto, J. Galvis, F. Moreno, and J. Vargas, “Identification of Super-Spreaders of Foot-And-Mouth Disease
in the cattle transportation network: The 2018 outbreak case in Cesar (Colombia),” Proc. 2019 IEEE World Conf.
Complex Syst. WCCS 2019, pp. 5–10, 2019, doi: 10.1109/ICoCS.2019.8930765.
[5] F. Basciftci and K. A. Gunduz, “Identification of Acidosis Disease in Cattle Using IoT,” UBMK 2019 - Proceedings, 4th
Int. Conf. Comput. Sci. Eng., pp. 58–62, 2019, doi: 10.1109/UBMK.2019.8907035.

842
[6] N. V. Kishan, S. T. Y, and S. Kavalur, “Cattle disease identification using Prediction Techniques,” no. 3, pp. 2764–
2770, 2021.
[7] E. M. Katemboh, R. Abdulla, V. Jayapal, S. K. Selvaperumal, and D. Ratnadurai, “Integrated animal health care using
IoT,” Int. J. Adv. Sci. Technol., vol. 29, no. 1, pp. 42–56, 2020.
[8] V. Garg and K. Garg, “Early Stage Disease Detection Platform in,” vol. 3, no. 11, pp. 13–15, 2016.
[9] L. Niu, C. Yang, Y. Du, L. Qin, and B. Li, “Cattle disease auxiliary diagnosis and treatment system based on data
analysis and mining,” 2020 5th Int. Conf. Comput. Commun. Syst. ICCCS 2020, pp. 24–27, 2020, doi:
10.1109/ICCCS49078.2020.9118590.
[10] D. Ashar, A. Kanojia, R. Parihar, and S. Kudoo, “Livestock Disease Prediction System,” VIVA-Tech Int. J. Res. Innov.,
vol. 1, no. 4, pp. 1–3, 2021.
[11] S. Shahinfar, D. Page, J. Guenther, V. Cabrera, P. Fricke, and K. Weigel, “Prediction of insemination outcomes in
Holstein dairy cattle using alternative machine learning algorithms,” J. Dairy Sci., vol. 97, no. 2, pp. 731–742, 2014,
doi: 10.3168/jds.2013-6693.
[12] A. D. Sunny, S. Kulshreshtha, S. Singh, M. Ba, and H. Sarojadevi, “Disease Diagnosis System By Exploring Machine
Learning Algorithms,” Int. J. Innov. Eng. Technol., vol. 10, no. 2, pp. 14–21, 2018.
[13] L. Wan, W. Bao, A. Disease, D. Expert, S. Based, and T. Ifip, “Animal Disease Diagnoses Expert System Based on
Long Wan , Wenxing Bao To cite this version : HAL Id : hal-01055402,” pp. 0–8, 2014.
[14] D. Wu et al., “Lameness detection of dairy cows based on the YOLOv3 deep learning algorithm and a relative step size
characteristic vector,” Biosyst. Eng., vol. 189, pp. 150–163, 2020, doi: 10.1016/j.biosystemseng.2019.11.017.
[15] H. Sadeghi, H. S. Braun, B. Panti, G. Opsomer, and O. B. Pascottini, “Validation of a deep learning-based image
analysis system to diagnose subclinical endometritis in dairy cows,” PLoS One, vol. 17, no. 1 January, pp. 1–14, 2022,
doi: 10.1371/journal.pone.0263409.
[16] V. Sturm, D. Efrosinin, M. Öhlschuster, E. Gusterer, M. Drillich, and M. Iwersen, “Combination of sensor data and
health monitoring for early detection of subclinical Ketosis in dairy cows,” Sensors (Switzerland), vol. 20, no. 5, pp. 1–
19, 2020, doi: 10.3390/s20051484.
[17] N. Misk, T. Misk, A. El-Khamary, and M. Semeika, “A retrospective study of surgical affections of mammary glands in
cattle and buffaloes and their management in the field,” J. Vet. Med. Sci., vol. 80, no. 10, pp. 1576–1583, 2018, doi:
10.1292/jvms.17-0586.
[18] A. Hidalgo, F. Zouari, H. Knijn, and S. van der Beek, “Prediction of postpartum diseases of dairy cattle using machine
learning,” Proc. World Congr. Genet. Appl. to Livest. Prod., vol. 11, no. February, p. 104, 2018, [Online]. Available:
http://www.wcgalp.org/proceedings/2018/prediction-postpartum-diseases-dairy-cattle-using-machine-learning.
[19] D. Y. Pavkin, A. S. Dorokhov, F. E. Vladimirov, I. M. Dovlatov, and K. S. Lyalin, “Algorithms for detecting cattle
diseases at early stages and for making diagnoses and related recommendations,” Appl. Sci., vol. 11, no. 23, 2021, doi:
10.3390/app112311148.
[20] Y. Ali, A. Farooq, T. M. Alam, M. S. Farooq, M. J. Awan, and T. I. Baig, “Detection of schistosomiasis factors using
association rule mining,” IEEE Access, vol. 7, pp. 186108–186114, 2019, doi: 10.1109/ACCESS.2019.2956020.
[21] M. Bhanu Sridhar, P. Aiswarya, and L. Kavitha, “Detection and Prediction of Frequent Diseases in India through
Association Technique using Apriori Algorithm and Random Forest Regression,” Int. J. Eng. Res., vol. V9, no. 03, pp.
386–393, 2020, doi: 10.17577/ijertv9is030427.
[22] R. Singh and E. Rajesh, “Prediction of Heart Disease by Clustering and Classification Techniques,” Int. J. Comput. Sci.
Eng., vol. 7, no. 5, pp. 861–866, 2019, doi: 10.26438/ijcse/v7i5.861866.
[23] A. A. Elngar, R. Kumar, A. Hayat, and P. Churi, “Intelligent System for Skin Disease Prediction using Machine
Learning,” J. Phys. Conf. Ser., vol. 1998, no. 1, 2021, doi: 10.1088/1742-6596/1998/1/012037.
[24] M. Shamsul Arifin, M. Golam Kibria, A. Firoze, M. Ashraful Amini, and H. Yan, “Dermatological disease diagnosis
using color-skin images,” Proc. - Int. Conf. Mach. Learn. Cybern., vol. 5, pp. 1675–1680, 2012, doi:
10.1109/ICMLC.2012.6359626.
[25] N. Wagner et al., “Machine learning to detect behavioural anomalies in dairy cows under subacute ruminal acidosis,”
Comput. Electron. Agric., vol. 170, p. 105233, 2020, doi: 10.1016/j.compag.2020.105233.
[26] W. Xu et al., “Prediction of metabolic status of dairy cows in early lactation with on-farm cow data and machine
learning algorithms,” J. Dairy Sci., vol. 102, no. 11, pp. 10186–10201, 2019, doi: 10.3168/jds.2018-15791.
[27] B. C. Hernández, N. Lopez-Villalobos, and M. Vignes, “Identifying health status in grazing dairy cows from milk mid-
infrared spectroscopy by using machine learning methods,” Animals, vol. 11, no. 8, 2021, doi: 10.3390/ani11082154.
[28] E. Afshari Safavi, “Assessing machine learning techniques in forecasting lumpy skin disease occurrence based on
meteorological and geospatial features,” Trop. Anim. Health Prod., vol. 54, no. 1, 2022, doi: 10.1007/s11250-022-
03073-2.
[29] N. Abdul Ghafoor and B. Sitkowska, “MasPA: A Machine Learning Application to Predict Risk of Mastitis in Cattle
from AMS Sensor Data,” AgriEngineering, vol. 3, no. 3, pp. 575–584, 2021, doi: 10.3390/agriengineering3030037.
[30] U I Bajwa*, S Alam, N u Haq, N I Ratyal and M W Anwar, “Skin Disease Classification using Neural Network”
Volume 16 , Issue 6 , 2020.

843