E-HEALTHCARE MONITORING USING MACHINE LEARNING

Sayon Chakraborty Shubham

Department of Computer Science Department
of Computer Science
Sharda University Greater Noida Sharda
University Greater Noida

Shreya Singh

Department of Computer Science

Sharda University Greater Noida

ABSTRACT

The patient data is classified using the classification approach. Many researchers pioneered
a wide range of healthcare techniques, from diagnostics to treatment. On the basis of an
effective e-health monitoring system, therapy and prevention may be provided however,
the degree of accuracy did not improve. In the sphere of portable and smart IoT devices,
the healthcare sector has seen significant advancement. The analysis of huge patient data
sets by clustering and classification of the data is one of the ways that has been employed to
offer this E-health monitoring system a new direction. Clustering is the process of creating
a comprehensive dataset of patients with a specific ailment and then analysing their various
data points. The data of the patients is then classified using classification. Many strategies
have been explored to enhance the E-Health monitoring system, but none of them has
shown to be effective in solving the problem. Many academics are analysing a large amount
of data on patients in order to remedy the problem with the health-monitoring technologies
using different machine learning approaches, adequate accuracy may be reached. Machine
learning techniques are significantly more efficient for patient health monitoring and
provide effective performance of “feature selection, classification, and clustering of data”
by sifting through a massive database of patient information.

Keywords: - Clustering of data, Machine Learning, Health monitoring systems, set of

algorithms, IOT devices

INTRODUCTION

Health monitoring system is one of the most important I set in the future for the identification
and prevention of illnesses like diabetes in nations like India, which has the highest number of
diabetic patients in the world, utilising big data and machine learning methods. One of the most
important innovations in the healthcare business is the usage of IoT Devices, which allows
patients' vitals to be measured in a cost-effective manner. There are a variety of health
monitoring systems available on the market. We're going through things like "activity tracking,"
"SpO2 level," "heartbeat monitoring," and so on. There is a large data set in the suggested system
via which it can anticipate.

The proposed system has a large data set through which it can predict the disease criteria on
which the patient will fall, as well as recommend a hilly diet and rate of exercise in the patient's
daily life. The goal is to make this system more and more accurate by analysing a large patient
data set. The first approach is to go through with the diabetes prediction and other diseases
prediction that the patient will fall into. The system will attempt to forecast the outcome with
increasing accuracy in order to provide a better consumer interactive experience. Including this it
will assists the patient in understanding what foods are required and what steps and diet should
be followed to be healthy. The system is analyzing the data that is available and through which it
can predicts the disease and what are the steps and diet should one follow to get rid of it or be
that in control. The available devices that are present in the market are basically based on IOT
but very of them are precise and accurate which will cost the customer’s pocket. Beside this the
devices need a lot of technical and not easy to use, The Nobel approach is to make a better user-
friendly E-health Monitoring system that is precise and cost effective. [1] In today's Internet
world, people are frequently confused about diseases, and a large amount of data on the internet
is frequently unable to effectively answer the question of a patient's symptoms due to a lack of
required dataset. In this proposed system, the symptoms are entered by the patient themselves,
and the system predicts which disease the patient may be suffering from based on that
information. The suggested system issue statement is based on the notion that individuals
nowadays are highly interested in learning what type of sickness they have, even if their
symptoms are moderate. We used to look for solutions all over the internet, but we never
received the ideal answer; instead, we got a lot of unwelcome results, which made us feel even
more anxious and afraid. So here is the proposed system, which is a machine learning based
python IDE based system that will receive the user's symptoms data and categorise it using its
many classification and other primary algorithm techniques to predict the patient's anticipated
disease. To make the prediction a little clearer and easier, it uses two training and testing
datasets. To make the prediction a little clearer and simpler to comprehend by the user, it uses
two training and testing datasets. The user may enter a set of five symptoms, and the system will
analyse which is the precise predictable disease for the provided collection of symptoms. There
are three main machine algorithms that are employed to increase the system's accuracy.

LITREATURE SURVEY

Using big data and machine learning approaches, a health monitoring system is one of the most
essential set in the future for the detection and prevention of illnesses like diabetes in countries
like India, which has the greatest number of diabetic patients in the world. The use of IoT
Devices, which allows patients' vitals to be measured in a cost-effective manner, is one of the
most essential breakthroughs in the healthcare industry. On the market, there are a range of
health monitoring systems to choose from. "Activity tracking," "SpO2 level," "heartbeat
monitoring," and other terms are being discussed. The recommended system has a vast data
collection from which it can forecast.

TITLE AUTHOR Software / Hardware ALGORITHM FINDING

Requirements Achieve

“Smart Health Monitoring and Mr. Ravish Gupta, Internet of things, Machine learning The hea
Management Using Internet of Dept. of Electronics Cloud Platforms, techniques, industry
and Artificial Intelligence, greatly
Things, Artificial Intelligence with Un-supervised
Communication Supervised the Inte
Cloud Based Processing” machine learning
Engineering (IoT) an
Convolutional Neural techniques
Intellige
Networks, Sensors and
Deep learning such bu
Actuators,
technica
Microcontrollers and
applicati
Electro-
wearab
Cardiogram. utilised
human
use of s
technol
expandi
the cha
and rea
diagnos

“STUDY ON AN IOT BASED Mr. Adars Internet of Things , Healthcare Wireles

SECURE” applications , sensor n
Vinayaka Mission’s Sensor network ,
be used
“HEALTHCARE MONITORING Semantic Web
Kadakkal, Kollam, patients
SYSTEM UNING technology,
Kerala healthc
CRYPTOGRAPHY” applicati
field for
sensor n
(WMSN

“REAL TIME HEALTH G Sahithi , P Iot-Internet of Things, Machine Learning A health

MONITORING USING IOT WITH Vinayasree , Pallati Monitoring, Sensors, approach. monitor
INTEGRATION OF MACHINE Narsimhulu Anatomy, Physiology, uses dig
Multiple Linear
LEARNING APPROACH” Cardiovascular and IoT
Regression
improve
Algorithm. Random
traditio
Forest Algorithm.
system'
 Support Vector informa
Machine collectio
method
analytic
Machin
algorith
program
languag
determ
likelihoo
disease
Using an
continu
monitor
was pro

“Development of Smart Md. Milon Islam , Internet of things · Creating a IOT based Internet
Healthcare Monitoring System in Sensors · health monitoring (IoT) tec
Ashikur Rahaman ,
IoT device with the help enables
ESP32,
Md. Rashedul develop
Environment” different sensors
Islam1 Pulse sensor, healthc
temperature sensor, to-face
BP sensor, ECG sensor, telemed
and raspberry pi. machine
are used
records
center s

“Analysis on E-Healthcare V. Deepa, K. Internet of Things, Use of IOT Big stati

Monitoring System with Iot and Rajeswari telemedicine, e-health interconnected with notably
Big Patient Data” the Artififcial nueral healthc
monitoring system,
networks , Deep medical
prevention, artificial
learning without
neural network
analyse
person
in a brie
Healthc
gadget
has acti
healthc
gadget
statistic
 Charact
selectio
and affe
categor
affected
statistic

“IoT based Health Monitoring Srivardhan Reddy IoT, Machine Learning Using of IOT devices The idea
System using Machine Learning” K1, Sidaarth R2, Sai to get the patient of Thing
Aneesh Reddy3, Dr. health diagnosis and Machin
Rajashree Shettar 4 through machine notably
learning analyzing inside th
the data. scientifi
healthc
to revea
circums
“Real Time Machine affected
Health Monitoring System Using Using o
Tzen Ket IOT Based Devices ,
Machine Learning and IOT” Iot based hardware anetwo
Wong ,Hou Kit Neural Network and encoded in the ware to
Mun, Swee King virtual machine artififcal nueral actual a
Phang, Kai Lok Lum
network and by using predicti
and Wei Qiang Tan
VM ware power is
given to the system

The system is evaluating the data available and predicting the ailment, as well as the measures
and diet that need be followed to get rid of it or keep it under control. The existing gadgets on the
market are mostly based on IOT, but few of them are exact and accurate, putting a strain on the
customer's wallet. Aside from that, the devices require a lot of technical knowledge and are
difficult to operate. The Nobel strategy is to create a more user-friendly, precise, and cost-
effective E-health Monitoring system. People are frequently puzzled by diseases in today's
Internet age, and a great volume of material on the internet is frequently unable to adequately
address the query.

Methodology
The proposed system, which is a machine learning based python IDE based system that will
receive the user's symptoms data and categorise it using its many classification and other primary
algorithm techniques to predict the patient's anticipated disease. To make the prediction a little
clearer and easier, it uses two training and testing datasets. To make the prediction a little clearer
and simpler to comprehend by the user, it uses two training and testing datasets. The user may
enter a set of five symptoms, and the system will analyse which is the precise predictable disease
for the provided collection of symptoms. There are three main machine algorithms that are
employed to increase the system's accuracy.

Figure 1. Workflow of the system

1. Workflow
The dataset for the proposed system is collected as two csv files that contain informative
information about various types of diseases and their severe symptoms, and by analysing the
different symptoms and using classifier techniques, the prediction of the particular disease can be
done. There are several machine learning prediction and classification techniques used to
perform the prediction of the disease, which increases the system's accuracy. The following are
some of the classification approaches that were utilised to create the prediction system:
1.1 Content Filtering Method:-In order to produce predictions, the content-based technique is a
domain-dependent algorithm that emphasizes analyzing the attributes of objects. When Pages,
articles, and news are presented as files, and Content-based filtering is without a doubt the most
profitable filtering method. In the event of an emergency, the recommendation is based on a
content-based filtering method. Characteristics obtained from data in the objects are used to
create human profiles. A person who has previously investigated
1.2 Collaborative based Filtering Method - Collaborative filtering is a domain-agnostic
prediction method for media such as movies and music that isn't well represented by metadata.
The collaborative filtering method saves user preferences for entries in a database (user-item
matrix). In a newer, narrower definition, collaborative filtering is a method of collecting
preferences or taste information from a large number of users in order to develop automatic
predictions (filtering) about a user's interests (collaborating). The collaborative filtering
technique's core premise is that if two persons A and B have the same viewpoint on one issue, A
is more likely than a randomly selected individual to share B's viewpoint on a different topic. [9]

a. Memory-based Filtering Method- The user's prior ratings are important in his hunt for a
neighbour who shares his interests. When a person's neighbour is recognised, a variety of
algorithms that take into account the preferences of friends may be utilised to provide
suggestions. These tactics have produced considerable success in real-world circumstances due
to their applicability.
b. Model-based Filtering Method - In this technique, models are constructed using various data
mining and machine learning methods to predict user ratings of unrated items. There are a
number of model-based CF approaches to choose from. Bayesian networks, clustering models,
singular value decomposition, probabilistic latent semantic analysis, multiple multiplicative
factors, latent Dirichlet allocation, and Markov decision process based models are all examples
of latent semantic models.
1.3 Hybrid Filtering Method - Memory-based and model-based CF algorithms are used in a
number of applications. Traditional CF techniques have limitations, and these methods increase
prediction performance. Importantly, they address CF concerns such sparsity and data loss.
However, the complexity and cost of implementation have grown. The majority of commercial
recommender systems, such as Google's news recommender system, are hybrid.
1.4 Naïve Baiyes - The strategy of creating classifiers in models to assign through a
representative vector value using a limited number of class levels is known as naive bias. There
is no other approach in which a large number of algorithms are based on the same premise as the
night bias classifier, which is considered to be independent of the features assigned to its class
variable. This is one of the most extensively used machine learning approaches for data
categorization and filtration, with applications in neural networks and heavy computing. [10]
1.5 Decision Tree - The decision tree is one of the most powerful and extensively used tools for
classification and prediction of data in machine learning. It appears like a tree-like structure and
creates a flow chart with the internal node and attribute asset, with each branch representing a
different type of data. Starting at the class level, this is the entry used to categorise the data, and
each node is its terminal node. By dividing them into three groups and leaving a comment, the
data is classified at the beginning of the tree node, and the attribute supplied by the node is
verified. The same method will be followed as you move up the tree branch based on its attribute
value. [11]

1. Implementation of the proposed system

In the proposed system, we use the Python IDE to import all of the training and testing
datasets, and we use various types of classifiers in the needed labels to make illness
 predictions using various types of machine learning approaches. The following is a step-by-
step guide to putting the system in place:

Step 1. Header Files Used

Figure 2. Header Files used

Pandas use for reading the dataset

Numpy used to convert features into numpy and then perform different further operations
accordingly.
Tkinter is used for creation of the interface of the proposed system.
K-Means is used for clustering.
Train_test_split was used to divide the dataset into train and test portions to train and test the
model.
Random Forest classifier is used to predict the the diseases based on the symptoms.

Step 2. Training and Testing the Data

 Figure 3. Training and testing of the Data

The data in the csv files that are to be used for training and testing of various diseases and their
symptoms are being implemented on various methods of classifiers, such as the random forest,
the naive bayes, and decision tree making. The data is being used to analyse the symptoms of
various diseases and predictions are being made on the basis of that. The set of csv files training
and testing helps to give outa a set of symptoms and a set of diseases that are present in the
current situation by analyzing and predicting measures we can determine that the confusion of
the people about any disease can be particularly resolved.

Step 3. Analysing the data using Classifiers and calculating accuracy

 Figure 4. Using Classifiers for Analysing

The data of the files will be tested by different classifiers techniques so that the prediction can be
done on that basis. Not only one classifier technique will be used if one fails, the other two will
respond accordingly. The use of the prediction system is necessary because people have been
confused between different diseases and their symptoms for a long time. This system not only
resolves that problem that most people have been having.

Step 4. Adding of GUI interface

The system must have a graphical user interface (GUI) through which the patient can enter his or
her set of symptoms. The patient simply needs to select from a list of symptoms that are already
listed in the symptoms set. Based on the symptoms, the prediction technique works, and with the
help of various classifier techniques, it can predict which disease is causing the patient's
symptoms.

Figure 5. GUI Interface

The GUI interface of the system have the part where the patient can enter its name and then the
set of 5 symptoms one patient can enter and on the basis of that set the prediction can be done by
the classifiers, one can also choose its desired prediction technique that may help in better
prediction but in most of the cases the techniques give out the same result in all cases. With the
help of Tkinter (python module used for graphical interface) there is a communication can be
generated between the patient and the the health monitoring system.
RESULT
After all of the above steps have been completed successfully, the result should be in the form of
a GUI interface created with Tkinter (a Python library for creating graphical user interfaces). The
prediction techniques used in this system are quite accurate in themselves and help to predict the
proper disease based on their training and testing dataset. The system can cluster the data and
then train and test the data using various techniques to produce the best prediction.

Figure 6. Result of the Health monitoring system

 Figure 7. Result in conflict of algorithm cases

There is a conflict between the data in the training and testing phases because the system is still
in the testing phase and the data set is much larger, so there is a chance that there might be
clashes or incorrect predictions, but there is also a thing that the patient must keep in mind that
the symptoms that are entered must be in appropriate order otherwise the prediction may come
out incorrect in conflict situation the patient can go with the tested two output of the classifier
techniques the accuracy is pretty good and there is a lot of scope of improvement can be done in
this field. The larger the data we have and better features of algorithms if used this system must
give out a plenty set of prediction that can be useful in the field of the health for the greater
cause.

Conclusion
The proposed system problem statement is based on the fact that people nowadays are very
curious to know what kind of disease they are suffering from, especially if they have mild
symptoms. We used to search all over the internet for answers, but we never got the perfect
answer; instead, we got a lot of unwanted results, which made us more panicked and scared. So
here comes the suggested system in light, which is a machine learning based python IDE based
system that will accept the user's symptoms data and classify it using its different classification
and other main algorithm approaches to anticipate the patient's presumed ailment. There are two
training and testing datasets used in it to make the prediction bit clearer and easier to understand
by the user the user can have a set of five symptoms which are to be entered by the patient and
on the basis of that the system can analyse it which is the exact predictable disease for the given
set of symptoms there are three different machine algorithms used so that the accuracy of the
system can be improved. There is a conflict between the data in the training and testing phases
because the system is still in the testing phase and the data set is much larger, so there is a chance
of clashes or incorrect predictions, but the patient must keep in mind that the symptoms must be
entered in the correct order or the prediction will be incorrect. In a conflict situation, the patient
can choose between the tested two outputs of the system.

Future Scope
The suggested method uses a vast data collection to anticipate which illness criteria the patient
would meet, as well as suggest a hilly diet and activity pace in the patient's everyday life. The
objective is to improve the accuracy of this system by analysing a huge set of patient data. The
first method is to proceed with the diabetes and other ailments that the patient is likely to
develop. In order to give a better customer interactive experience, the system will attempt to
anticipate the outcome with increasing accuracy. The future scope of the proposed system is
quite intriguing; if the proposed system can be connected to a set or a cloud of datasets with a
large amount of training and testing data, then the outcome must be much more precise;
additionally, the algorithms that are being used must be more specific and predictive, or to put it
another way, more accurate; therefore, the outcome must be quite interesting. Becoming anxious
as a result of online searches for their symptoms and receiving strange replies If we talk about
the system's GUI interface, the system uses a python-based GUI called Tkinter, which is pretty
old fashioned. The system can be connected to the frontend either through API to communicate
with users in a better way, or by giving a better platform so that people can live a hassle-free life
and get nervous less by surfing and searching for their symptoms and getting strange answers
from the internet. When it comes to the system's graphical user interface, the system employs a
Python-based GUI called Tkinter, which is quite outdated. Aside from that, there is a lot of work
that can be done in the field of system deployment and maintenance to keep track of every
patient on the internet by providing a better platform so that people may live a stress-free life and
feel less scared when browsing.is using a pythoin based GUI Tkinter that is preety old fashioned
the system can be connected to the frontend either by help of APIto communicate with user in a
better way.

Acknowledgement
With profound respect, I express my deepest appreciation to Dr..Mandeep Kaur, Assistant
Professor in the Department of Computer Science Engineering at Sharda University, for his
support & cooperation in the completion of this study. Working with him has taught me a lot,
and We will always be grateful for his contribution to my life.
References
[1] Yong-Yuan Deng, Chin-Ling Chen, Woei-Jiunn Tsaur, Yung-Wen Tang, and Jung-Hsuan
Chen, “Internet of things (IoT) based design of a secure and lightweight body areanetwork
(BAN) healthcare system”, Sensors, vol. 17, no. 12, pp. 2919, 2017.
[2] Mr. Ravish Gupta, Dept. of Electronics and Communication Engineering Smart Health
Monitoring and Management Using Internet of things Artificial Intelligence Cloud based
processing 2017.
[3] Mr. Adars , Vinayaka Mission’s Kadakkal, Kollam, Kerala study on an iot based
securehealthcare monitoring system using cryptography 2017.

[4] G Sahithi , P Vinayasree , Pallati Narsimhulu real time health monitoring using iot with
integration of machine learning approach 2019.
[5] Md. Milon Islam , Ashikur Rahaman ,Md. Rashedul Islam Development of Smart Healthcare
Monitoring System in IoT

Environment 2018.
[6] V. Deepa, K. Rajeswari Analysis on E-Healthcare Monitoring System with Iot and Big
Patient Data 2017.
[7] Srivardhan Reddy K1, Sidaarth R2, Sai Aneesh Reddy3, Dr. Rajashree Shettar IoT based
Health Monitoring System using Machine Learning 2019.
[8] Tzen Ket Wong ,Hou Kit Mun, Swee King Phang, Kai Lok Lum and Wei Qiang Tan
Real Time Machine Health Monitoring System Using Machine Learning and IOT 2018.
[9] Tzen Ket Wong , Hou Kit Mun , Swee King Phang, Kai Lok Lum and Wei Qiang Tan Patient
health monitoring using IoT with machine learning 2020.
[10] Aaditya Jain, and Bhupendra Kumar Soni, “Secure modern healthcare system based on
internet of things and secret sharing of IoT healthcare data”, International Journal of Advanced
Networking and Applications, vol. 8, no. 6, pp. 3283, 2017.
[11] Prabhu M, Seethalakshmi G and Gollapudi Anisha, “Secured healthcare system in IoT”,
International Journal of Pure and Applied Mathematics, vol. 118, no. 20, pp. 3239-3244, 2018.