5th International Conference on Electrical Engineering and Information Communication 

Technology (ICEEICT), 18-20 November, Dhaka

Application of Machine Learning in Predicting

Individual and Collective Health System

Department of Electrical, Electronic and Communication Engineering

Military Institute of Science and Technology
  Dhaka, Bangladesh
[email protected], [email protected] 

Abstract—The current health system is highly dependent on in patients regarding the hospital's readmission, approaching
manual procedures. Patients have to be highly dependent on sepsis in the stay of the hospital, and exasperated state of a
the doctor’s appointments and advice. Machine learning medical disorder. Healthcare organizations demand to
performs a significant role in accomplishing prediction models
concentrate on estimating their complete organizational
and data gained from diverse sources to manage enormous
systems like healthcare. Basically, it’s an implementation of a
tactics while guaranteeing the redisposition of AI for
prediction model about the healthcare industry which will take determining particular issues like the deception of hospital
part in improving the overall efficiency. By gathering a readmissions. The subsequent action lies in designating
database and categorizing the data into particular categories funds for transforming the IT substructure so that there is AI
will help to execute this work thoroughly. Using machine predisposition. This substructure encompasses various and
learning’s basic principle, a systematic prediction of diseases far-reaching incorporations of secure data for qualifying the
will be provided and this will ease up the recovery process of models of ML without consideration of their visible
the patients very quickly. Simultaneously, the patients will purchases or inward development. The most effective
receive a proper guideline to suppress the disease they have to
application of ML in the healthcare division is in
cooperate with their body’s overall health. For chronic disease
cases, the patients will get various benefits from it. The distinguishing and recognizing complications or ailments
proclamation of the whole theme will be executed by the neural and their medications that are, on the other hand,
network methodology of the machine learning algorithm. unmanageable to determine. It can be comprehensive of
cancers that are challenging to discover.
Keywords— Machine Learning, Deep Neural Network, During beginning platforms, by efficiently employing
SVM patients' medical records, redesigned medications and
prescriptions can be improved by ML technologies for
I. INTRODUCTION earmarking distinct diseases among different patients.
The confluence of multiple areas of knowledge has Consolidating portentous analytics with machine learning
commenced in the configuration and implementation of can be meaningful in healthcare [3]. Health care field
computer structures in the health context. At present there operates a huge amount of information about clinical
are various innovative health systems implemented with AI assessment, patient condition report, cure of particular
techniques where machine learning algorithms play a role disease, regular follow-ups, medicine and so on. It is a very
primordial. AI is a new and promising sector of computer complex matter to put together properly. The inappropriate
science. It can design models based on the activities of a management of information has created an impact in the
particular being. These activities include two fundamental quality of data alliance or association. Improvement and
components: rationalizing or reasoning and behavior. reformation in the measure of data needs some authorized
Machine learning is a quintessential part of AI, which aims way to assemble and process information very accurately.
to evolve methods that let computers acquire knowledge. Some classifiers can differentiate between data formulated
[1]. on the components is authenticated by one of the main
The health industry needs to overcome the challenge machine learning applications.
of computerized data management by subsuming it The client's health obligations enhance the
effectively into combing all the AI techniques. This accomplishment of healthcare with big data and machine
challenge has had several attempts in the past couple of learning. Electronic health document entreaties are being
years [2]. The preponderance of the applications of ML is escalated in this current circumstance, which needs to
within the early stage of research in the entire healthcare concentrate on utilizing the data produced by those
sector. ML on Radiology is the maximum researched arena statements. There is a comprehensive measure of healthcare
in the healthcare sector. Researchers are examining the data associated with unconventional practices, principally
exercising of ML for distinguishing diseases and other neurological and cardiac [4].
health contingencies through digital imaging. Pertinency of The paper manifests a record examination of research
ML is unmistakable in numerous areas like evaluating risk that is captivating deep learning in health informatics. It
decorates a significant analysis and consulting the applying algorithms every system component’s behavior
corresponding advantage of its future viewpoints. The paper is analyzed and predicted.
essentially presents the meaningful applications of deep
B. Deep Neural Network
learning in overflowing biomedical data, radiological
imaging techniques, computational techniques and public It’s one of the most used architectures of deep learning.
health. And the method used here is primarily deep learning It is basically used for classification. It has more than 2
[5]. hidden layers as the system requires. It can also process
Research contributed specific knowledge concerning complex or non-linear hypotheses.
the devices in the healthcare region. The accessories C. Support Vector Machine
accommodate composing more dependable determinations
A support vector machine (SVM) is a supervised
established on the past congregated data with the help of an
algorithm for two-group classification problems. After
exceptional process mechanism analyzing with big data
then giving an SVM model sets of labeled training data
techniques. They used the mechanism of producing
for each category, they’re able to categorize new text.
possessions like Machine Learning and Artificial
Intelligence with healthcare within a sole plan of action.
And for the research approach, they used “Machine
Learning Mechanism” in Collecting Past Healthcare Data III. PROPOSED MODEL
[6].
Another research aims to use various techniques of Firstly, the data from various hospitals are collected.
machine learning and determine how these models can help After the validation and removal of the outliers of the data,
us to solve health problems. They used State of the art the residual data were interpreted using ML algorithms or
machine learning, Analysis of different data repositories, procedures that foretell laboratory test outcomes. Data will
Transformation and normalization. And different algorithms be collected from various sources.
like Knn, Naive Bayes, Random Forest and SVM were used.
Where Random Forest and SVM algorithms showed better Health Data
results than Naive Bayes and Knn algorithms. The arena of
health informatics (HI) endeavors to provide a wide-
reaching interconnection among different approaches.
Usually, a healthcare dataset is found to be inadequate and
incomplete. So in the field of software engineering reading
from the dataset is actually a failure. Numerous ML Classified
Non-Classified
mechanisms can be applied to dissolve and examine data
that can encourage the character of performance for both 1. Medical ID
1.Laboratory
medical and pharmaceutical staff as well as doctors [7]. 2. Age
Results
In the domain of deep learning, a maintained 3. BMI
2.Diagnosis
convergence of health informatics analysis around deep 4. Disease
Notes
learning ultimately could consign to second place the 5. Symptoms
3. Images
advancement of current machine learning algorithms with a
numerous comprehensive use of computational support and
accountability.
As for future work, to continue on in this area of AI as
it seems very interesting to us and to propose other studies
Collected data will be divided into two portions:
with other data sets to see the behavior of the classification
classified and non-classified. Data will be collected from
algorithms. We also consider making use of other systems
individuals, hospitals, clinics, diagnostic centers, health
for AI like "R" to be able to compare results. We would also
departments etc. Then the data will be processed.
like to use one of the ASUM or CRISP-DM methodologies
[8].

The remnants of this paper are constructed as follows.

Section 2 presents the associated technology like behavior
modelling, deep neural network, support vector machine etc.
Section 3 describes the proposed model based on deep
neural networks with different machine learning algorithms.
Section 4 emphasizes detailed Performance evaluation of
proposed models and finally Section 5 concludes the paper.

II. RELATED TECHNOLOGY

A. Behaviour Modeling
This model is used to describe the behavior of
every system component. Using all the collected data
and
 Fig. Flowchart of the Whole Process

Ultimately, the whole process will predict the health

condition of individuals and gathering them a will provide a
detailed result of a certain area.

IV. PERFORMANCE COMPARISON

Fig: Proposed Model Framework The present system does not emphasize the
Our proposed model will work like this. By providing implementation of the cloud domain. The plan discusses two
some questionnaires to the consumers of those hospitals to different things, which are to control health care with
collect some insightful data. We put these two kinds of data reliability and give the consumers the level of satisfaction
into a machine learning model by analyzing it with the they need on this basis of health care. In their time of
machine learning algorithms. This analysis correlates with vulnerability, the cloud system will guide them to make
original datasets and impersonates familiarities from the fruitful decisions.
data applying ML algorithms to develop the effectiveness of If the sufferer is begetting any problem and if there is no
the requisite system.  opportunity of petitioning for cooperation or notifying the
doctor of the caregiver, then utilizing the machine learning
Regression, clustering, preprocessing, classification, algorithm, an IoT device which is combined to the cloud
dimensionality reduction, deep neural network algorithms will necessitate the features of the health of the sufferer and
are used to analyze the datasets and after that, the system the data which was acknowledged as the data will be
chooses the perfect model to do further work. confirmed by the machine learning models as that is
Our processed data will be categorized in clinical data accurate or not for the assurance. If the information is
and behavioral Data. After encoding the data, different genuine, then the erudition will be concocted, and the
machine learning algorithms will be applied. After the physician will understand the sufferer and forecast the
model selection the algorithm will do testing, evaluating, results. [4].
training and then prepare and read the final data. Cost effectiveness will be another premise for the
consumers by implementing the work. Hospitals will know
   every little detail of the medications they are providing to
the consumers and they can keep track of the scarcity of the
medications properly.

IV. CONCLUSION
Our health system is definitely getting better day by day,
but as of now it isn’t automated as much as it needed to be.
People still have to wait in the queue, they still have to do
the tests and submit the vitals over and over. Plus, they don’t
have a proper system from which they can get exact
suggestions regarding maintaining their health. Machine
Learning can save both our health and precious time by
managing and predicting the data and processing it. So, by International Conference on Computing Methodologies and
Communication (ICCMC), 2019, pp. 1211-1215, doi:
using our knowledge we need to implement machine 10.1109/ICCMC.2019.8819782.
learning in our health system to make it more efficient. [4] G. Sasubilli and A. Kumar, "Machine Learning and Big Data
Again it will be a great help for maintaining the budget of Implementation on Health Care data," 2020 4th International
the pharmaceutical industry on the basis of making the Conference on Intelligent Computing and Control Systems (ICICCS),
medicines with higher demand. This will actually protect the 2020, pp. 859-864, doi: 10.1109/ICICCS48265.2020.9120906.
wastage of medicine in the third world countries. [5] D. Ravì et al., "Deep Learning for Health Informatics," in IEEE
Journal of Biomedical and Health Informatics, vol. 21, no. 1, pp. 4-21,
Jan. 2017, doi: 10.1109/JBHI.2016.2636665.
[6] R. Singh and N. Sharma, "Machine Learning based Medical
Information Analysis, Estimations and Approximations over Present
REFERENCES Health Research Domain," 2020 International Conference on
Computational Performance Evaluation (ComPE), 2020, pp. 704-708,
[1] D. Cedeno-Moreno and M. Vargas-Lombardo, "Application of doi: 10.1109/ComPE49325.2020.9200045.
Machine Learning with Supervised Classification Algorithms: In the [7] M. Moreb, T. A. Mohammed and O. Bayat, "A Novel Software
Context of Health," 2019 7th International Engineering, Sciences and Engineering Approach Toward Using Machine Learning for
Technology Conference (IESTEC), 2019, pp. 613-618, doi: Improving the Efficiency of Health Systems," in IEEE Access, vol. 8,
10.1109/IESTEC46403.2019.00115. pp. 23169-23178, 2020, doi: 10.1109/ACCESS.2020.2970178.
[2] K. R. Dalal, "Analysing the Implementation of Machine Learning in
Healthcare," 2020 International Conference on Electronics and
Sustainable Communication Systems (ICESC), 2020, pp. 133-137,
doi: 10.1109/ICESC48915.2020.9156061.
[3] D. Dahiwade, G. Patle and E. Meshram, "Designing Disease
Prediction Model Using Machine Learning Approach," 2019 3rd