DESIGN AND ANALYSIS OF HEART ATTACK PREDICTION SYSTEM USING MACHINE LEARNING ALGORITHM

Abstract

Any diseases that is kept untreated for a long period of time directly impacts the human health and effect all the major organs of the body. Treating the chronic diseases in timely basis is highly recommended to avoid major medical problems coming over the way. On implementing Chronic Heart Attack prediction system using Hybrid SMLT (Split Multi Link trunking) echnique is presented here. The proposed system considers machine learning algorithm in which Logistic regression (LR), multi-level perceptron (MLP) based on neural network, CatBoost regression (CB) algorithm and Random Forest regression algorithm (RF) is comparatively tested. Considering the data set provided, the presented approach is implemented using Python IDE and simulated in loT (Internet of things) screen of Google Collab. A comparative analysis of various Machine learning algorithm is helpful to make the accurate processing with the help of robust method in short span of time. Further different machine learning techniques, with multiple features are considered as scope of implementation. The proposed approach is creating a machine learning model that can predict whether or not a person will experience a heart attack. The best model is used to forecast the result after many techniques have been compared

Specification

Myocardial infarction (MJ) is Cardinal necrosis is evaluated from persistent
coronary changes which are normally called as myocardial infarction. The
disease get version by chronic symptoms until experiencing a heart attack
people may not realise the presence of disease with them. Various symptoms
and relief are provided by the detection of myocardial infarction and further the
performance of daily activities greatly reduces the infection. Necessity
expensive test are suggested such as PPG, electrocardiogram (ECG) sensors are
helpful to analyse the pattern of Heart function. The radiation of data collected
from the heart is helpful to analyse t)le pattern of signal using algorithm that is
being evaluated. The result shows effective ECG based method for detecting
myocardial infarction with similar accuracy. Cardiovascular diseases create
millions of death cases every year it is one of the serious problems in recent
days. The primary reason for death due to cardiovascular disease is due to the
lack of early warning people experience symptoms that need to be treated
immediately as an emergency condition. Various research frameworks are implemented usmg machine learning algorithms such as random for a
struggression linear aggression artificial neural network K nearest algorithm
etc .. In order to predict the heart diseases in the yearly stages Technology
related support is highly helpful. The majority of clinical diagnosis is based on
data collected from the patience on the regular activities with medical support.
Predicting the hard disease in the early stages is helpful to treat the disease
effectively and further improve the quality of living. The clinical information
collected from mining the data of patient regular activity and hybrid the data
into intelligent algorithm to provide the prediction system. The presented
approach considers machine learning algorithm to analyze the heart disease fat
and by collecting dataset. Depending on various symptoms that related to the
cardiovascular diseases' treatment plants are based on the severe of the disease
present. The cardiovascular diseases are depended upon various Lifestyle habits
such as changes in life style habits tobacco products altering the regular diodes
etc. The medications such as cardiovascular disease related medicines are based
on the physical condition and historical health conditions. Surgery are
procedures are provided by the doctors depend upon the medications
undertaken. Cardiovascular diseases are initially treated as a serious problem in
recent days . The heart is one of the important functions. Active monitoring of
cardiovascular disease infected Patience is highly important hence automated
monitoring systems are evaluated in recent days because of the technology
development.
4.2 FIELD OF INVENTION
The invention is to develop a machine learning model for heart attack Prediction,
to potentially replace the updatable supervised machine learning classification
models by predicting results in the form of best accuracy by comparing
supervised algorithm.
The proposed approach is focused on implementing exploratory data analysis
enabled classification model using Hybrid Split Multiple Link trunking
(HSML T) technique.
• The input data is collected from clinical information of normal and chronic
heart problem affected patients. The data is pre-processed, by removing the junk
values present in it.
• Cleaning the data removes the empty attribute columns present in it. Further the
data is split into training data and testing data.
• By applying Machine learning algorithms such as logistic regression, multiLevel
perceptron, Random Forest regression, CatBoost regression techniques, the
comparative performance is analysed here. • The goal is to develop a machine learning model for heart attack Prediction, to
potentially replace the updatable supervised machine learning classification
models by predicting results in the form of best accuracy by comparing
supervised algorithm.
4.3 DISCUSSION OF THE RELATED ART
Some previously completed research presented a method that discuss about
various machine learning algorithms to predict chronic heart disease from a
binary and multiple classification perspective. Using Novel decision tree
algorithm, the proposed method investigated here with the concurrent location
of various heart disease relativity data. To differentiate the bags of frameworks
(BOF) with more reliable performance in terms of accuracy, this Decision Tree
(DT) framework enhances its performance. The proposed methods are
developed with binary classification and multi-class classification are hybrid
together. Another research presented a novel system that provides comparative
analysis model with significant features for the early identification of
cardiovascular disease (heart disease). The prediction technique employs
reliable non-classification techniques to provide the novel variations. The
commonly utilized machine learning algorithms are provided such as support
vector machines (SVM), logistic regression model (LR), random forest model
(RF), and linear regression algorithm (LRA). In spite of accurately detecting
(CHD) Chronic Heart disease, the presented study discussed hybrid model is
presented here. The proposed method focused on improving the features
extracted from the considered standard dataset, achieved with relative accuracy
of 88. 7%. An Internet of Things (loT) framework for creating deep convolution
neural network (DNN) towards heart disease identification system. Another
presented a Novel approach, where prognostic approach is used here to show
the semantic understanding of stroke forecast framework. The discussed
technique considers heart problems and its side effects by considering ECG and
PPG data. The novel approach considers data from old patients. For early stroke
prediction, the author presented a strategy focuses on existing pattern of heart
care data, and created a model with a convolution neural network (CNN) and a
long short term memory network (LSTM). The author provides the method for
utilizing physiological data from the patient and formulated into biosensors and
wearable sensors. On the other hand, the novel approach considers 97.51%
accuracy of the early-stage prediction of heart disease and stroke is considered
here. Strokes are considered as the serious issue in elderly people, where the
problem can occur at any time.
4.4 SUMMARY OF INVENTION
The suggested approach is creating a machine learning model for categorizing
myocardial infarction. The procedure starts with data gathering, when historical information on heart disease is gathered. If the cardiac condition is identified
early enough, lives can be saved. Data cleansing and visualization for
diagrammatic representation are the first steps in the analysis process. Data
analysis is carried out on the dataset with the correct variable identification,
which results in the discovery of both dependent and independent variables. The
dataset where the data pattern is learned is then subjected to the correct machine
learning algorithm. A more effective algorithm is utilized to anticipate the
outcome after experimenting with several methods
4.5 DETAILED DISCRIPTION OF THE INVENTON:
A. Data Pre-processing :
The proposed model is started with machine learning (ML) based comparative
analysis model, validation methods are used to determine the difference in error
rate, in spite of different dataset with regards to actual error rate. Validation
methods are not required always if the data collection is large enough since
medicine data are normally big data. Moreover, in actual conditions, the real
working with data samples is not accurate in many cases where it represent the
dataset strength. In order to discover the duplicate value, empty spaces, missing
data, and numerous data type, unique float variable is integer. The sample dataset
is used to have the pattern, to tune model existing hyper parameters and provide an
independent pattern of evaluation in terms of relativity with dataset.
B. Data visualization :
The study of machine learning various techniques are employed to analyse the
collected input data in order to find the pattern present in it. The quality of analysis
process is estimated using statistical measures evaluated using the formulas such
as main square error rate accuracy Precision sensitivity etc. The quantitative
analysis and estimators are helpful to have a primary focus of statistical findings.
The qualitative comprehension with data visualisation provides significant result
of tools. The visualisation process has helpful to identifY the patterns to check the
character data to make outline of the data and to explore more about the unknown
data set. Data visualisation is a helpful to express and demonstrate relativity
between each plot and further have a decision making towards domain knowledge .
It is recommended to drive deep are into the systems to make data visualisation
and explorer data analysis effectively
C. Logistic Regression
Logistic regression data collected from the data set is being classified using
various machine learning techniques. The data preparation is implemented before
fetching the data into any analysis module. The given data is considered 70% for
training data set 30% for testing data set. On the model is created a part of the testing data is switched as a real time input which is helpful to predict the output.
Some of the machine learning algorithms implemented here are discussed below.
Logistic regression it is a statistical technique for analysing unknown data to find
the outcome of one are more independent variables. The Logistic regression
employees to possible research to be measured. It is also called as a binary
regression. In Logistic regression the best model to evaluate relationship between
each set of independent data are variable is a formulated using the response from
the variable. The probability of category assigned by the dependent variable is
employed to make the Logistic regression. The dependent variable in Logistic
regression is binary value with unknown data it is being coded as one for
successful regression and zero for failure.
D. MLP Classifier
Multilayer perceptron classifier is used to make robust analysis of unknown
dataset in which multiple layers is employed connected with the neural networks.
MLP classifier provides neural network as the base layer for making the decision
and further the classification will support factor machine etc. The major factor of
MLP classifier and other classification algorithms from scientific learning is that in
common it is required to have no more effort on implementing other machine
learning models since the layers are integrated within the library itself so through
Python programming language it accesses a plug and play model initially this
required to configure the layers using the library installation that's all.
E. Random Forest Algorithm
Random forest regression random forest regression is also called as randomly
associated collective decision making in which at the training of the data set large
trees of decisions are constructed in each trees have a specific output. Random
decision for rest has the tendency of eliminating the decision trees to over fit the
training data set. The ensemble approach of a random for a cell are them enable the
type of supervisor learning method further it is useful to make a decision on
multiple angles. The ensemble approach calculus various decisions and various
levels of decisions to make the final outcome. In case of issues with intermediate
decision model The Final Decision will be affected. Random forest algorithm is
considered as one of the robust methodologies for making analysis on unstructured
the data set. Figure 3 shows process steps for RF algorithm.
D. CatBoost Classifier
Cat boost classifier algorithm are category based boosted decision tree
algorithm is one of the techniques in machine learning methods. It is function as
similar way as the XGBoost algorithm and gradient boost algorithm is performed.
It supports decision making based on category-based regression approach. It
supports high accuracy without tuning the parameters and support graphical
Processing Unit (GPU) of the system to have optimised training process.Split my feeling trunking advanced method of multiple link trunking that
provides enhanced benefit of a bandwidth allocation in aggregation. It avoids the
failure of multiplexing problems occurring over the network the existence of
SMLT technique in machine learning provides computers to provide continuous
analysis of input data with a different pattern. HSML T is advanced form of link
aggregation scheme where the capability of selecting different frameworks is being
utilised tax as a multiply sir where the switches will be connected to the required
channel based on the best performance in the similar way the proposed to approach
consider Logistic regression algorithm and the forest channel and father select the
best one from the results in terms of accuracy.
I Proposed On Chronic Heart Attack prediction system using Hybrid SML T (Split Multi Link
trunking) technique.
The proposed system considers machine learning algorithm in which Logistic regression (LR), multilevel
perceptron (MLP) based on neural network, CatBoost regression (CB) algorithm and Random
Forest regression algorithm (RF) is comparatively tested.
Considering the data set provided, the presented approach is implemented using Python IDE and
simulated in loT (Internet of things) screen of Google Collab.

Documents