Predictions And Analytics In Healthcare: Advancements In Machine Learning

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 07 | July 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 235
Predictions And Analytics In Healthcare: Advancements In Machine
Learning
Vangala Surya Teja Reddy1, Gatla Akanksh2, Satyam3, Kalpana Kumari4, Bhanu Talwar5
1234Undergraduate student, Lovely Professional University, Jalandhar, Punjab
5Professor, Lovely Professional University, Jalandhar, Punjab
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - As everyone in this worldknowshowtechnologyis
advancing every day and the drastic changes are occurring in
every sector these days and so does in the healthcare sector.
Many revolutions are coming ourwayfromthissectorbecause
the integrated technology in this sector is helping the
scientists, researchers, doctors, etc. to reach their target goals
with the help of computers. They can get the exact match or
near to that with the help of technology and in the same way
the user can also be informed and be filled with theknowledge
with the help of these technological advancements. Like, in
today’s world, we are seeing a lot of websites or apps that are
helping users know their diseases by getting some input from
them and they are also getting proper medication based on
that result. In this review, there is information about
predictive analytics that is framing the healthcare sector and
making it more comfortable for the end-user to diagnosetheir
diseases and take charge of them in minimal time. This paper
is going to talk about various advancements that can lead to,
what we say as, informed user experience, which will keep
them informed and also, they will be abletocurethemselvesat
an appropriate time.
Key Words: Healthcare sector, technological
advancements, diseases, predictive analytics, diagnose,
user experience.
INTRODUCTION
It is very essential for the diagnosis of chronic diseasesinthe
medical field as these diseases persist for a long time. Some
of the most common chronic diseases include diabetes,
strokes, heart disease, arthritis, cancer, hepatitis C. If wecan
detect the disease in an early phase, it will help in taking
preventive actions and effective treatment at an initial stage
which is always found helpful for patients[1] in curing their
diseases.
There has been lots and lots of data that can help the
healthcare sector to improve on many weak points and also
to do different advancements through the use of modern
technologies such as machine learning,artificial intelligence,
data mining techniques, and many more. These latest
technologies combined can open many new doors for the
people in this world and can help scientists and doctors to
reach a particular milestone in a veryshortperiod.Thereisa
lot of potential that needs to be unlocked because the things
which were to be done in years are now possible within a
fraction of seconds if all the things are placed in order. So
basically, it can improve the efficiency of the human brain
and also be a part of a tremendous changeofera withtheuse
of technology. In this research paper, a problem that needs
to be solved is presented so that further development in this
area can be made, and also with the help of existing
technologies, something can be improved by using a
tremendous amount of data that is available to us. There is
always a question in people’s minds about how a particular
disease occurs and how to be able to cure that in the first
place. So, by appropriately using some technology, we got to
know that now people can assure their well-being by sitting
at their homes, and also if they got to know something bad
about them, then they can cure it as soon astheyknowabout
it. We are going to predict some of the diseases for them to
let them know whether they are having any chances of
catching that disease. Moreover, users’ data will beanalyzed
about various implications or factors relatedtogettingsome
diseases at a particular time, age, etc.
LITERATURE SURVEY
As we think it is easy to predict or analyze data, but it is not,
it is certainly not so easy to analyze big-data data[4]withinit
and there are too many challenges to predict something to
the end user in the right way.
The difficulty of analyzing big data comes from its three
dimensions, namely, variability, speed, andvolume.‘Variety’
suggests that large amounts of data are made up of many
types of data, both formal and informal, e.g., doctor notes.
Healthcare data is presented in a variety of formats and
presentations from a variety of sourcesincluding(1)Clinical
data on Electronic Health Records, medical imaging,
machine-sensory data, and genomics, (2) clinical and R&D
data e.g., from clinical trials. and journal articles, (3) job
claims and cost data from health care providers and
insurance companies, and (4) patient behavior and
emotional data from wearable devices and communication
posts, including Twitter feeds, blogs, Facebook status
updates, healthcare communities, and web pages. ‘Speed’
refers to large data that is transmitted and obtained in real-
time, e.g., critical signals, and usually arrives at an explosive
rather than continuous level. ‘Volume’ means large data, in
the name itself, is extremely large. For example, 3D CAT
scanning usually takes 1 GB while one human genome takes
about 3 GB of data[2]. The role of professionals in thefield of

health care should maintain a balance betweenbusiness and
technology, which paves the way for Big Data in healthcare
information systems. Using Big Data in a health care
environment has led to changes in the storage and handling
of complex, randomized data sets[3].
Several researchers have been committed in recent years to
evaluating the classification precision of various
classification algorithms used in the Cleveland heartdisease
database [9], which is freely availablethroughtheUniversity
of California's online data mining repository. Many
researchers have used this database since its inception to
investigate various classification challenges using various
classification algorithms. Detrano [10] deployed a logistic
regression approach and achieved a classification accuracy
of 77%.
The author of [11] researched on the Clevelanddata setwith
the goal of analyzing global computational intelligence
techniques and found that applying a new approach
improved prediction performance. The performance of his
proposed technique, on the other hand, is dependent on the
qualities chosen by the algorithm.
P. Saranya's paper[7] has mentioned a different algorithm
which is used for predicting accurate results after diagnosis.
Machine decision diagnosis auxiliary algorithm is used to
diagnose cancer with large datasets, this algorithm has 77%
accuracy in predicting the disease. An online contextual
learning algorithm is used because it minimizes the false
positive rate in breast cancer screening diagnosis decisions.
Also, Multiple Streams of 2D convolution networks are used
for detecting false positives in the scan. When 3D
Convolutional Neural Network isuseditperformsbetter asit
has a sensitivity of 85% to 90% in detecting false positives.
Gamification, Convolutional Neural Network algorithm, and
Online contextual learning algorithm is used for predicting
outputs as accurately as possible.
Dhiraj Dahiwade, Gajanan Patle, and Ektaa Meshram
proposed a generic disease prediction system[8] based on
the patient's symptoms. For disease prediction, researchers
used the K-Nearest Neighbor (KNN) and Convolutional
Neural Network (CNN) machine learning algorithms. The
CNN algorithm has an accuracy of 84.5 percent in general
disease prediction, which is higher than the KNN approach.
In addition, KNN has a higher time and memory need than
CNN. After predicting general disease, their system can
calculate the risk of general disease, indicating whether the
risk is lower or higher.
Min Chen, Yixue Hao, Kai Hwang, Lu Wang, Lin Wang, In this
proposed paper[12], They streamline techniques for
effective chronic disease outbreak prediction in disease-
prone communities. They modified prediction models using
real-world hospital data from central China between 2013
and 2015. They employed a latent component model to
recreate the missing data to overcome the challenge of
incomplete data. And experimented on a regional chronic
disease of cerebral infarction. Using structured and
unstructured data from hospitals, they suggested a new
convolutional neural network (CNN)-based multimodal
disease risk prediction algorithm. When compared to
various common prediction algorithms, their new approach
has a prediction accuracy of 94.8% and a convergencespeed
that is faster than the CNN-based unimodal disease risk
prediction algorithm.
TECHNOLOGY USED
Various technologies like AWS, EC instance; tableau, Flask,
Pandas, Numpy, and Scikit Learn have been used to
implement this project and to analyze and predictthevalues
of the users, the data is converted into meaningful
information to let users experience the advancements in
technology.
In this implementation to predict and analyze the data,
different libraries of python(like pandas, NumPy, Scikit
Learn) are being used for data processing,data visualization,
and model building.
Numpy and Pandas have libraries that are used for any
scientific computation, here we used this due to their
intuitive syntax and high-performance computation
capabilities.
Numpy: It is used to provide support for large
multidimensional array objects like shape manipulation,
logical and mathematical operation, sorting, selecting, Basic
algebra & statistical operations, and much more. In Numpy
we can perform element-wise operations on your dataset. It
provides us with an enormous range of fast and efficient
ways of manipulating data inside them. Numpy uses less
memory to store data and work on arrays, which is very
convenient to use. During Data processing we use Numpy's
function like sort() for Sorting elements, to add elements we
use concatenate() function. Wecandoindexingandslicingof
the data according to our requirements.
Pandas: It is an open-source python package that is mostly
used for data analysis and machine learning tasks. Pandas
allow us to do time-consuming and repetitive work in a very
simple way. This work includes Data cleaning, data fill, Data
normalization, data visualization, statistical analysis, Data
inspection, loading & saving data, and many more. Pandas
are used as one of the most importantdata manipulation and
analysis tools. Pandas allow importing from different file
formats such as comma-separated-values, JSON, SQL
database tables, and Microsoft Excel.
Scikit-learn: It is the most useful and robust library of
python used for machine learning. Scikit-learn allows us to
define ML algorithms and evaluate many other algorithms
against one another. It also includes tools to help us
preprocess our datasets.Thispackageprovidesa selectionof

efficient tools for machine learning and statistical modeling
including classification, regression, clustering, etc. The real
power of Scikit-Learn lies in its model evaluation and
selection framework, where we can cross-validate the
parameters of our models. With the help of this library, we
can create a machine learning model that will predict the
positive or negative output model. This library is focused on
modeling data with the help of Ensemble methods, Feature
extraction, feature selection & Supervised models.
Flask: Flask is a web framework of Python modules that let
us develop web applications easily. It has a small and easy-
to-extend core. Flask comes undera micro-framework,it has
little to no dependencies on external libraries.
AWS: Amazon Web Services (AWS) provides you with
trustworthy, scalable, and cost-effective computational
resources for any applications. AWS offers a massive global
cloud infrastructure that enables us to quickly create, test,
and iterate. Rather than waiting weeks or months for
hardware, users can rapidly deploy new apps, scale up as
their workload expands, and reduce resources as demand
decreases. That's why AWS is preferable for web hosting.
METHODOLOGY
A single algorithm may not be able to make the perfect
prediction for any given dataset. As a result of their
limitations, machine learning algorithms are not able to
produce accurate models.
Accuracy counts all of the truepredictedvalues,however not
specific for each label that exists. To completelyremovethis,
we use various algorithms.
If we build and combine multiple models, the overall
accuracy can be increased. The combination can be
implemented by aggregating the outputs from different
models with two objectives: reducing the model error and
maintaining its generalization.
Some techniques can be used to implement such
aggregation.
Fig 1: Multiple model combination
The ensemble[5] model construction did not focus solely on
the variance of the algorithm used. In this case, each model
learns a specific pattern specialized in predictingoneaspect,
which is known as a weak learner, and those weak learners
may use different strategies to map the features with
different decision limits.
Bagging: Using bagging, training data is made available for
an iterative learning process. Each model learns the error
that is produced by the previous model using a slightly
different subset of the training datasets. Bagging helps us to
reduce the variance and overfitting.
Fig 2: Bagging methodology
Boosting: Adaptive Boosting is an ensemble of algorithms,
where we build models on the top of several weak learners.
As a result of such adaptability, sequential decision trees
adjust their weights based on prior knowledge of accuracy.
Hence, we perform the training in such a technique in a
sequential rather than parallel process.Inthistechnique, the
process of training and measuring the error in estimatescan
be repeated for a given number of iterations or when the
error rate is not changing significantly.
Fig 3: Boosting methodology
Gradient Boosting: With high predictive performance,
gradient boosting algorithms are great techniques.
Stacking: Stacking is analogous to boosting models; they
produce more robustpredictors.Stackinginvolvesbuildinga
stronger model from all weak predictions made by weak
learners.
Decision tree: A decision tree (DT) is one of the earliest and
most prominent machine learning algorithms. A decision
tree models the decision logic, i.e., how to classify data items
into a tree-like structure based on the results of tests. The
nodes of a Decision tree normallyhavemultiplelevelswhere
the first or top-most node is known as a root node. All
internal nodes represent tests on input variables or
attributes. According to the results of the test, the branches
of the separation algorithm point to the correct location of
the child where the process of testing and integrating
reaches the leaf position again. Terminal nodes are the
outcomes of the decision process. A common component of
medical diagnostic protocols, DTs are easy to interpret and
simple to learn.

Random forest: A random forest (RF) is an ensemble
classifier[6] consisting of many DTs similar to the way a
forest is a collection of many trees. Random subsetsarebuilt
from the original dataset. To identify the appropriate split,
each node in the decision tree considers a random set of
features. A decision tree model is fitted to each subset. All
decision trees are averaged together to determine the final
prediction.
DTs that are grown very deep often cause overfitting of the
training data, resulting from a high variationinclassification
outcome for a small change in the input data. Their training
data is quite sensitive, which causes them to make errors on
the test dataset. Training datasets use differentpartstotrain
the different DTs of an RF. The input vector of a new sample
is used to classify it. For a sample to be classified, it must
pass down to each DT in a forest. Each DT then The
classification is based on a different part of the input vector.
The forest then selects the classification based on the most
votes (for discrete classification outcomes) ortheaverage of
all trees in the forest (for numeric classification outcomes).
In addition to considering the results of many different DTs,
the RF algorithm can reduce the results of a single DT of the
same dataset.
RESULT
We are just beginning to understand what can be done with
Big Data in the healthcare industry today. Big Data will be
able to generate innovative and novel solutions to the
problems of healthcare as a result of the intersection of data
from multiple sources, tools, and technologies. There are a
lot of implementations that can be achieved using a massive
amount of data that every part of healthcare is collecting
daily from its customers, types of diseases, and many more.
Likewise, there are a lot of advancements going on in the
healthcare industry but there are lots and lots of
advancements that need to be introduced in it to let people
get interested in their health and their loved ones. And for
the same purpose, it has been developed to let people
understand themselves better with the help of Machine
Learning and Artificial Intelligence which is going to
revolutionize this. It issolvingmanymoreproblems soeasily
in a fraction of a second with its implementations.
In this, predictions have been made on some diseases for
people by taking some values from the user and giving them
a proper result based on their inputs. Through this, it is
going to get a step closer to predicting diseases for the user.
Here are some samples of how it has taken inputs and how
you are going to get output based on the information that is
provided in the columns.
Fig 4: Breast cancer UI
In the above image, as you can see, user can predict their
chances of getting breast cancer by providing certain values
like the Mean of the concave points, the Mean of thearea,the
Mean of the radius, etc. By calculating these values they are
going to get output telling them their chances of having
breast cancer as seen in the below image.
Fig 5: Positive result
Fig 6: Lung cancer UI
If someone wants to predict their chances of having lung
cancer then, as shown in the above image, they can also
predict it by adding certain values asked like Albumin,
Albumin and Globulin Ratio, Total Proteins, etc. which will
allow them to predict their disease in an early phase. And
also the user will get the output based on the information
provided as you can see in the below image.

Fig 7: Negative result
After predicting some of the diseases, we have taken the
same diseases to analyze the data of that disease. We have
used Tableau and Python to analyze the data which is being
fetched from Kaggle itself. Analysis of the data can also be
modified by making some adjustments for which we have
included some toggle buttons and sliders which will allow
the user to adjust the ranges according to which they can
know about the data that they want. For their convenience,
some labels have been included to guidethemtounderstand
the graphs and they will be having certain values based on
the ranges and data available at the moment. For example,
how many pregnant ladies can have the risk of catching
diabetes on an average whether it isBenignorMalignant ata
particular age? There is a lot more information that can be
consumed from the graphs and people can educate
themselves by taking a look at those graphs. Here are some
samples of these graphs which a person can see after
selecting a particular disease on the webpage.
Fig 8: Diabetes analytics UI
In the above image, it is shown that, between the age of 21-
81, with a glucose level of 0-199, blood pressure between 0-
122, and based on many more factors, how many average
pregnancies can catch diabetes, which can be benign(Green
bar) or malignant(Red bar) and it is also shown for the
glucose levels of diabetic patients at that point.
Fig 9: Heart disease analytics UI
In the above image, the graphs are being shown for having
heart disease. There are various factors which are affecting
for the type of chest pain that a personcanhavewhilehaving
heart disease in the above graph it is having Thalach range
between 71-202, Cp range of 0-3, Trestbps range of 94-200,
Fbs between 0-1 and many more ranges are being used to
analyze this data. It is also shown in the graph about who all
are safe and who can be affected by giving accurate numbers
on an average using different ranges and inputs.
CONCLUSIONS
In today’s world, a lot more diseases can affect people in
many ways but there are very less technologies that can be
accessed and used by the user itself to self-educate or to get
some information on this serious topic. Many technological
geeks are working in the direction of educatingpeopleabout
the health of themselves and their family members. If a user
is educated and has access to the information which is
appropriate and precise then it is easy to diagnose the
diseases and also people can get alertness or consciousness
in themselves which can help them get treated before it gets
worse.
So, through this project, which is backed by properresearch,
we want to get ahead with the help of technology in
predicting people’s health by using MachineLearningand AI
and also analyzing some data for them so that they can gain
some knowledge through this website. The sole motive of
this project is to give people the freedom to live and take
care of their health and we who are here have a little bit of
interest in technology and want people to know about what
Artificial Intelligence can do for them and how it would be
framed for our future selves. In this project, we have used
some latest technologies to build a web application that can
show your chances of catching some diseases by getting
some values from the users. It is a prediction model which
can show information about your disease by asking you for
some information. And in the second part, we have analyzed
some datasets which will be havingdynamicvaluesbywhich
one can know about a particular disease, e.g., at what age
there are more chances of getting breast cancer or which
types of things will let you catch a heart disease. There will
be graphs to show this information to the user.

In this, we have leveraged the amount of data that is
available out there which can be made of use if someone
knows how to use it properly. We have tried to make this
information meaningful so that these new technologies can
be introduced to the people and can be of use in any way.
REFERENCES
[1] H. Alharthi, “Healthcare predictive analytics: An
overview with a focus on Saudi Arabia,” Journal of
Infection and Public Health, pp. 749–756, Feb. 2018.
https://doi.org/10.1016/j.jiph.2018.02.005
[2] Divya Jain, Vijendra Singh, “ Feature selection and
classification systems for chronic disease prediction: A
review”, Egyptian Informatics Journal,pp.179-189,Mar.
2018. https://doi.org/10.1016/j.eij.2018.03.002
[3] Atreyi Kankanhalli, Jungpil Hahn, Sharon Ta, Gordon
Gao, “Big data and analytics in healthcare: Introduction
to the special section”, Business media Newyork, Mar.
2016. https://link.springer.com/content/pdf/10.10
07/s10796-016-9641-2.pdf
[4] A.Rishika Reddy, P. Suresh Kumar. “ Predictive Big data
analytics in healthcare”, Second International
Conference on Computational Intelligence
Communicational Technology (CICT), Feb. 2016.
https://sci-hub.hkvisa.net/10.1109/CICT.201 6.129
[5] Mohammed Alhamid, “A guide to learning Ensemble
technique”, Toward Data Science, Mar. 2021.
https://towardsdatascience.com/ensemble models-
5a62d4f4cb0c
[6] An introduction to Random forest Classifier.
https://scikit-learn.org/stable/modules/gen
erated/sklearn.ensemble.RandomForestClassifier.html
[7] P. Saranya, Dr. P. Asha, “Survey on Big Data Analytics in
Health Care”, 2019 International Conference on Smart
System and Inventive Technology(ICSSIT), Feb. 2020.
https://ieeexplore.ieee.org/document/8987 882
[8] Dhiraj Dahiwade, Gajanan Patle, Ektaa Meshram,
"Designing Disease Prediction Model Using Machine
Learning Approach",2019 3rd International Conference
on Computing Methodologies and Communication
(ICCMC), Aug. 2019.
https://ieeexplore.ieee.org/document/8819 782
[9] “Cleveland Heart Disease Dataset,” accessed on 03-02-
2017.
[10] R. Detrano, A. Janosi, W. Steinbrunn, M. Pfisterer, J.-J.
Schmid, S. Sandhu, K. H. Guppy, S. Lee, and V. Froelicher,
“International application of a new probability
algorithm for the diagnosis of coronary artery disease,”
The American journal of cardiology, pp. 304–310, 1989.
https://www.ajconline.org/article/0002-914
9(89)90524-9/pdf
[11] B. Edmonds, “Using Localised ‘gossip’ to structure
distributed learning”, Jan. 2005.
https://www.researchgate.net/publication/
28764359_Using_Localised_'Gossip'_to_Str
ucture_Distributed_Learning
[12] Min Chen, Yixue Hao, Kai Hwang, Lu Wang, Lin Wang,
“Disease Prediction by Machine Learning Over Big Data
From Healthcare Communities”, pp 8869 - 8879, April
2017. https://ieeexplore.ieee.org/abstract/docum
ent/7912315

Predictions And Analytics In Healthcare: Advancements In Machine Learning

More Related Content

Similar to Predictions And Analytics In Healthcare: Advancements In Machine Learning (20)

More from IRJET Journal (20)

Recently uploaded (20)

Predictions And Analytics In Healthcare: Advancements In Machine Learning