DL (Diagnostic Logistic)

Ankita Singh Khushi Tyagi Lavish Sadh

1900330100039 1900330100120 1900330100127
[email protected] [email protected] [email protected]
Raj Kumar Goel Institute Raj Kumar Goel Institute Raj Kumar Goel Institute
of Technology of Technology of Technology
Ghaziabad Ghaziabad Ghaziabad

Manan Singh Ms. Vernika Singh

1900330100129 Assistant Professor (Mentor)
[email protected] [email protected]
Raj Kumar Goel Institute Raj Kumar Goel Institute
of Technology of Technology
Ghaziabad Ghaziabad

Abstract Various analysts believe that insight cannot be

produced without learning. Machine learning
In today’s world, everyone deals with a variety of techniques come in a variety of forms, including
ailments as a result of their choices and unsupervised, semi-supervised, supervised,
surroundings. So, it becomes hard to make reinforcement learning, evolutionary learning, and
predictions deep learning. These lessons are used to classify
vast amounts of data extremely thoroughly.
The hardest duty is dealing with sickness. Each
year, there is a significant data increase in the This paper’s goal is to examine or predict diseases
medical sciences. The accurate analysis of on the basis of different symptoms.
medical data, which has benefited from early
patient care, has increased as a result of the rise of Therefore, we employ SVM (support vector
data in the medical and healthcare fields. Data machine) and K-Nearest Neighbor (KNN). due to
mining uncovers hidden pattern information in the medical
vast volume of medical data with the aid of
disease data. We suggested broad based on the
patient's symptoms, and diagnose the condition. K Data is increasing daily; therefore, using it to
Nearest Neighbor (KNN), Support Vector accurately anticipate disease is an essential duty.
Machine (SVM), and Random Forest algorithm of However, processing huge data is also essential in
machine learning algorithms are used to general, so data mining plays a significant role.
accurately forecast disease.

Machine learning makes it simple to classify huge

datasets. Understanding how to diagnose patients
Introduction correctly through clinical examination and
assessment is essential. Inadequate information
The person's lifestyle choices and test results are
management has had an impact on the quality of
taken into account for an appropriate diagnosis in
the data association.
this general disease prediction. After predicting
general disease, this approach is able to indicate A legal method must be found to focus and
whether the risk of general disease is lower or process information in a viable and effective
higher. manner as data volumes increase. In order to build
a classifier that can separate the data depending on
Computers are now more intelligent thanks to
different criteria, one of the various machine
machine learning, which also gives them the
learning software is used.
ability to think.
The data set is divided into two or more classes.
These classifiers are used to analyze medical data
and forecast diseases. Today, machine learning is
pervasive, and one may utilize it frequently
throughout the day without even realizing it.

Methodology
1) K-Nearest Neighbour Algorithm
K-Nearest Neighbour is one of the simplest
machine learning algorithms, based on the
supervised learning method.
o the K-NN algorithm assumes that the new and
existing cases are comparable, and it places the
new instance in the category that is most like the
existing categories.
o the K-NN algorithm saves all the information
2) Support Vector Machine Algorithm
that is accessible and categorizes fresh data based
on similarity. This means that by utilizing the K- It is a supervised algorithm. Support Vector
NN method, fresh data can be quickly and Machine, or SVM, solves Classification and
accurately sorted into a suitable category. Regression problems. It largely works on Machine
Learning Classification issues. The SVM
o the K-NN approach can be used for both
algorithm's motive is to establish a decision
classification and regression problems, but it is
boundary that can divide n-dimensional space into
more frequently utilized for classification issues.
classes to classify fresh data points in the future. A
As K-NN is a non-parametric technique, it doesn’t hyperplane is the optimal decision boundary.
make any assumptions about the underlying data.
SVM chooses the extreme points that aid in the
It is also called as a lazy learner algorithm. It
creation of the hyperplane. Support vectors, which
saves the training dataset rather than learning from
are used to represent these extreme instances, are
it immediately. KNN uses the dataset to provide
what give the Support Vector Machine method its
result when classifying data.
name. Consider the diagram below, where a
the KNN algorithm simply saves the dataset decision boundary or hyperplane is used to
during the training phase and classifies fresh data categorize two distinct hyperplanes:
into a category that is quite similar to the training
The example we used for the KNN classifier can
data.
be utilized to understand SVM.
Example: Let's say we have a picture of a creature
Take an example that correctly distinguishes
that resembles both a cat and a dog, but we're not
between a cat and a dog, let's say we observe an
sure which one it is. We may therefore utilize the
unusual cat that also resembles a dog. We can
KNN method for this identification since it is
build such a model by utilizing the SVM
based on a similarity metric. Our KNN model will
algorithm. Prior to testing it with this weird
look for features in the new data set that are
animal, we will first train our model with several
comparable to those in the photographs of cats and
photographs of cats and dogs so that it can
dogs, and based on those features, it will classify
become familiar with the various attributes of cats
the data as belonging to either the cat or dog
and dogs.
group.
3) Random Forest Algorithm
Literature
Random forest is one of the preferred and
commonly used algorithms by data scientists. Chronic diseases are a major problem in the
Random forest and other supervised machine healthcare industry everywhere in the world. The
learning algorithms are frequently used in medical statement claims that chronic diseases are
classification and regression problems. It builds to blame for the rising death rate among people.
decision trees from different samples, using their Over 70% of the patient's income is spent on the
average for categorization and majority vote for disease's therapies. Therefore, reducing the
regression. patient's danger of passing away is really crucial.
The development of medical research facilitates
the acquisition of health-related data. The patient's
Steps Involved in Random Forest Algorithm medical history, demographic information, and
results of medical analyses are all included in the
Step 1: - In the Random Forest model, each healthcare data. Depending on the geographies
decision tree is constructed using a subset of data and the types of environments in those places, the
points and a subset of features. Simply, m features diseases that result could vary. Therefore, in
and n randomly chosen records are obtained from addition to the disease information, the data set
a data set with k records. should include information about the patient's
Step 2: For each sample, a unique decision tree is environment and residence.
built.
Step 3: An output will be produced by each The integration of information has accelerated the
decision tree. evolution of the healthcare sector in recent years.
Information Technology is a part of it. The goal of
Step 4: For classification and regression, the final integrating IT into healthcare is to improve
result is evaluated using the majority vote or people's quality of life by making it more
averaging. comfortable and inexpensive, similar to how cell
phones did so. This might be making healthcare
intelligent, such as the development of the smart
ambulance, smart hospital infrastructure, and so
on, which benefits patients and doctors in various Y. Zhang, M. Qiu, C.-W. Tsai, M. M. Hassan, and
ways. A. Alamri [4] created a cloud-based health-based
Cps system that manages the enormous volume of
IM.Chen, Y. Ma, Y. Li, D. Wu, Y. Zhang, and C.
biological data. Y. Zhang talked about the medical
Youn gives a concept for a wearable 2.0 system
industry's rapid growth in data volume. The issue
where
with large data is that it is created in a short period
Create intelligent washable clothing that enhances of time and has the characteristic of being stored
the Qu and QoS of the upcoming healthcare in various formats. The health-CPS system was
system. A New IoT-based data-collecting system created using two technologies, the first of which
was created by Chen. The innovative sensor-based is a big data technology and the second of which
smart home created a washable cloth. The doctor is the cloud. Numerous operations on cloud-like
used this cloth to record the patient's physiological data analysis, monitoring, and prediction were
state. Further analysis is done with the aid of the carried out using this system. One can learn more
physiological data. With the aid of this about how to handle and manage the enormous
component, the user is able to gather data on the volume of biological data on the cloud with the
patient's physiological state as well as their aid of this technology. Data collecting, data
emotional health status by using a cloud-based administration, and data-oriented layers are the
system. This inversion of washable smart fabric is three layers that the system takes into account.
mostly made up of numerous sensors, wires, and The layer that collected the data did it in a certain
electrodes. standard format. the parallel computing and
distributed storage data management layer
It was able to capture the patient's physiological
state with the aid of this cloth. And this data is With the aid of the health-cps system, 8 different
utilized for the analysis. discussed the difficulties operations are carried out by this system.
encountered when creating the wearable 2.0 Additionally, the vast array of healthcare services
architecture. that this system is aware of this.

The gathering of physiological data, detrimental L. Qiu, K. Gai, and M. Qiu discussed how to
psychological impacts, opposition to wireless manage a vast volume of hospital data in the cloud
body area networking, and sustainable massive and offered a telemedicine system. The author of
physiological data collection are among the this research recommended improvements to the
problems with the current system. the various telehealth system, which are mainly focused on
activities carried out on files, such as data the cloud-based data exchange amongst all
analysis, monitoring, and prediction. Once more, telehealth services. However, there are other
the author divides the functional elements of problems with cloud data sharing, including
Wearable 2.0's smart clothing into the following network bandwidth and virtual machine switching.
categories: sensor integration, electrical-cable- For better data sharing using data sharing
based networking, and digital modules. Numerous concepts, a cloud-based solution to data sharing is
applications are covered in this, including those recommended in this. With the use of this model,
for tracking chronic diseases, caring for the the author focuses on temporal restrictions,
elderly, and managing emotions. network capabilities, and transmission probability.
For this, the author created a brand-new, ideal
method for sharing massive data. Users are
B. Qian, X. Wang, N. Cao, H. Li, and Y.-G. Jiang provided with the best method for processing
[2] using the patient's EHR data, a risk prediction biomedical data by this algorithm.
system for Alzheimer's disease was developed.
Here, they used an active learning setting to
address a genuine issue the patient was
experiencing. The risk of Alzheimer's disease is
taken into account in that active risk prediction Proposed work
system
As we've shown in this paper, we can descry B. Comparison Results
conditions using Machine literacy with the help of
This section presents the performance of
symptoms. There are different criteria we can
the KNN, SVM, and Random Forest
descry the needed result similar to the system for
classification algorithms on the basis of
prognosticating conditions, where we developed
time required and
the top-position process of prognosticating
accuracy. Shows accuracy Comparison of
conditions. It provides a simple summary of the
all three
entire system. It's intended to give a quick
algorithms for various thresholds. We
overview of ails, croakers, and reports while
found out the SVM gives more accuracy as
displaying the system as a single, high-position
compared to KNN and random forest
process, along with its connections to cases and
algorithm
croaker External realities.
* Managing all the Cases
* Managing all the Symptoms
* Managing all the conditions
* Managing all the Croakers
* Processing Case records and induce reports of
all Cases.
* Processing Medicines records and induce
reports.

Confusion matrix for KNN on Test data

Result discussion
A. Experimental Setup
1) Run in a system environment using
a Windows 10 (64-bit) PC with an Intel
Core i5-6200U running at 2.30 GHz and
8GB of RAM.
2) Dataset
Patient dataset retrieved from GitHub.
Conclusion Common., vol. 55, no. 1, pp. 54–61, Jan.
We suggested a broad machine learning 2017.
algorithm- grounded complaint
vaticination system. We used KNN, SVM, [4] Y. Zhang, M. Qiu, C.-W. Tsai, M. M.
and random forest algorithms to categorize
case data since, in the ultramodern medical Hassan, and A. Alamri, “Health PS:
world, medical data is expanding greatly Healthcare cyber physical system assisted
and needs to be reused in order to give
by cloud and big data,” IEEE Syst. J., vol.
precise complaint prognostications
grounded on symptoms. By furnishing the 11, no. 1, pp. 88–95, Mar. 2017.
input of case records that let us understand [5] S Mohan, C Thirumalai, G Srivastava
the position of illness threat vaticination,
Effective heart disease prediction using hybrid
we were suitable to produce an accurate
general complaint threat vaticination. machine learning techniques. IEEE Access,
Because of this technique, disease and risk volume 7. Posted: 2019
prediction may be accomplished with little
effort and expense. We compare the [6] heart disease prediction and classification
outcomes of these algorithms in terms of
using machine learning algorithms optimized
accuracy and processing time. The
accuracy of the SVM algorithm is higher by particle swarm optimization and ant colony
than that of the other two algorithms, and optimization. Int. J. Intell. Eng. Syst, volume
the SVM processing time is lower than
12, issue 1. Posted: 2019
that of the other two algorithms. So, in
terms of accuracy and timing, SVM is
[7] A Mir, S N Dhage. 2018 Fourth
superior.
International Conference on Computing

References Communication Control and Automation

(ICCUBEA), p. 1 – 6. Posted: 2018
1] M. Chen, Y. Hao, K. Hwang, L. Wang,
and L. Wang, “Disease prediction by [8] M Mannerizing, M J Rahman, B Ahammed,

machine learning over big data from M M Abedin. Classification and prediction of

healthcare communities”,” IEEE Access, diabetes disease using machine learning

vol. 5, no. 1, pp. 8869–8879, 2017. paradigm. Health Information Science and
Systems, volume 8, issue 1. Posted: 2020
[2] B. Qian, X. Wang, N. Cao, H. Li, and
Y.-G. Jiang, “A relative similarity-based
method for interactive patient risk
prediction,” Springer Data Mining Knowl.
Discovery, vol. 29, no. 4, pp. 1070–1093,
2015.

[3] IM. Chen, Y. Ma, Y. Li, D. Wu, Y.

Zhang, and C. Youn, “Wearable 2.0:
Enable human-cloud integration in next
generation healthcare system,” IEEE