Introduction

At recent times, diseases related to heart is predominantly one of the main reasons for fatal and unexpected deaths that occur. A healthy lifestyle and primary finding are eminent and safest idea to prevent heart disease. The exertion projected herein focuses primarily on noticing the heart disease using machine learning approach with the help of medical test parameters. Human heart functions throughout the life span of a human being and considered to be the utmost vital part and any abnormality in its working develop a life-threatening health problem. Aberrations in heart is the basis for illness in other parts of the body which in turn is classified as can be classified as heart disease. Heart complications triggered due to insalubrious routine, smoking, liquor consumption and high cholesterol may cause high blood pressure [1]. The World Health Organization has a record where in around 10 million succumb due to heart disease per annum. A hale and hearty lifestyle and initial finding is the solitary means to avoid heart disease. The biggest challenge in health care today is delivering the best possible quality service and an accurate diagnosis. Even if still heart disease is considered as the leading cause of death, they are also the ones that can be managed and controlled effectively. The aim of this research paper is to test a patient for cardio related health issues by examining the on medical attributes such as gender, age, chest ache, abstaining blood sugar etc. A record of collected data is selected from the UCI source accompanied by the patient’s medical past and characteristics. Taking this dataset into account, we foresee if the patient, in near future faces heart related complications or not. With 76 attributes in hand only 14 attributes are considered for testing. These medicinal features are skilled under two algorithms: logistic regression and Random Forest Classifier. Random Forest which gives us the accuracy of 90.16% can be considered as the most efficient algorithm than the former.

Literature Survey

A quite substantial amount of work related to diagnosis of cardio vascular disease using machine learning algorithm has motivated this work.

Golande and Pavan Kumar studied on three different machine learning algorithms such as Decision tree, K-means clustering and KNN for classification of heart disease and found that accuracy produced by Decision Tree was best [2].

Alotaibi, carried out research to predict heart disease by implementing five different machine learning algorithms [1]. Their study used the Rapid miner tool which gave the highest accuracy than any other tool. The only drawback in their research was the size of the data set. From their analysis it was concluded that Decision tree produced the highest accuracy.

Lutimath et al. in their research worked on “Prediction Of Heart Disease using Naïve Bayes classification and SVM”. In their research SVM with radial kernel produced better accuracy than Naïve Bayes classification. Mean Absolute Error, Sum of Squared Error and Root Mean Squared Error were the major parameters used in the performance analyses [3].

Rajdhan and Sai and many others worked on the same using four machine learning algorithms [4]. Their research contrasts the exactness of Decision Tress, Naïve Bayes, Logistic regression and Random Forest and of the four Random Forest algorithm was the most effective. The research included metrics such as Accuracy, Precision, Fmeasure and Recall to carry out the performance analysis which was applied on pre-processed dataset.

Machine learning algorithms partake effectual in producing results with a high level of accuracy. It has aided health scholars and medics all over the world in recognising patterns in the patients ensuing in initial detections of heart diseases [5].

Results verified with hybrid blend of PSO with SVM (PSO_SVM) completes superior predictive performance over additional replicas. The study will consequently allow doctors to detect cardiovascular diseases and pledge apt action without the intrusion of a qualified cardiologist [6].

Proposed Model

Our projected work forecasts heart disease by discovering two classification algorithms (mentioned above) and ensures performance. The objective of this study is to detect heart disease using machine learning approach with the help of medical test parameters.

A healthcare employee inputs the values from patient’s health report which is fed into the model. The model using the data that is uploaded predicts if the patient is prone to heart disease or not using machine-learning algorithm. Figure 1 shows the complete process involved.

Fig. 1
figure 1

Proposed model predicting heart disease

Full size image

Dataset Assembly and Preprocessing

The data set used is a cardiology data set available in UCI Cleveland. This dataset is a collection of 76 attributes, out of which only 14 are well-thought-out for testing and even vital to substantiate the performance of above considered algorithms [7]. A detailed report of the 14 attributes specified in the proposed work can be found in the Table 1.

Table 1 Attribute description of UCI-ML Heart Disease Dataset Cleveland heart disease dataset is extracted from UCI-ML repository (Machine Learning Repository of University of California, Irvine; 2017) to comprehend the effect of hybrid feature selection and classification algorithms [8]
Full size table

Classification

The attributes listed in Table 1 are provided as input to various ML algorithms such as random forest, Logistic Regression Classification technique. The dataset is divided as training (80%) and test (20%) data sets. The training dataset is the data set used to train the model. A test dataset is used to validate the performance of trained model. The performance of each algorithm is calculated and analysed using various indicators such as accuracy [9]. Various algorithms considered in this paper is as follows.

Logistic Regression

Despite its name, logistic regression is more of a classification model than a regression model. Logistic regression is a simple and efficient method for binomial and linear classification problems [10]. Data mining algorithms in hybrid mode outperforms algorithms when used individually in diagnosing cardiovascular diseases. This is a classification model that is very easy to implement and achieves very good performance on linearly separable classes. This is a classification algorithm widely used in the industry. The output of logistic regression can take values such as yes or no, 0 or 1, true or false.

Random Forest

Random Forest, a commonly chosen supervised machine learning algorithm is used for both classification and regression. A classifier such as random forest algorithm considers a dataset of decision trees over diverse subclasses of a given dataset and norms them to improve the prediction perfection of these dataset [11]. Decision trees are built by random forest diverse subsets takes mainstream vote for classification and average in case of regression.

Summary

  • Classifiers like Support Vector Machines (SVM), Artificial Neural Network (ANN), Naïve Bayes, K-Nearest Neighbour (KNN), Random Forest and C5.0 Decision Tree are assumed to achieve improved results mutually with Particle Swarm Optimization (PSO) feature selection method.

  • UCI-ML heart disease dataset consists of 76 parameters over-all. Yet, mainstream of the available researcher and research papers mentions to using a subset of 14 attributes.

  • Accurateness, sensitiveness and specific metrics are the most common performance metrics for assessing models in spotting cardiovascular illnesses.

Analysis

The outcomes received by smearing Random Forest and Logistic Regression algorithms are depicted here. The Accuracy score is used to measure algorithm performance according to the values obtained. We use the recorded dataset and carry the experiments by implementing the above cited algorithms [12].

Our project projects us that Random Forest accomplishes good results than Logistic Regression Classifier by predicting whether the patient is diagnosed with a heart disease or not. This evidences that Random Forest is better in diagnosis of a heart disease [13]. The following Figs. 2, 3, 4, and 5 depicts a plot of the number of patients that are been separated and foreseen by the classifier reliant on the Chest Pain [14].

Fig. 2
figure 2

The figure shows how we obtain the accuracy score of heart disease using the algorithm [14]

Full size image
Fig. 3
figure 3

Target count of logistic Regression(0) and Random forest(1) [14]

Full size image
Fig. 4
figure 4

Severity of chest pain denoted as cp in the above diagram [14]

Full size image
Fig. 5
figure 5

Accuracy score achieved using Logistic Regression is: 85.25%. The accuracy score achieved using Random Forest is: 90.16% [14]

Full size image

Conclusion and Future Enhancements

Cardiac related disease is presumed as the main cause of demise amongst patients with illness. From our study, we can conclude that there is a necessity to innovate an automated system which predicts heart disease precisely [15]. The key challenge which motivated the learning is to choose the effective algorithm under ML for prediction of cardiovascular disease. This exertion correlates the correct notch of Logistic Regression and Random Forest algorithms in detecting heart disease as early as possible. From this study it is clear to consider the Random Forest algorithm as the utmost effective algorithm showing 90.2% accuracy rate in finding the heart disease [16]. Future enhancements, we can develop an online application that better displays the details of the patient along with the indication of possible heart attack. We can also reduce the number of parameters considered to predict the heart disease.