Cardiovascular disease classification and analysis of the risk factors using Machine Learning models
by
Embedded Files
PHASE I: INTRODUCTION
Introduction
Introduction
Cardio and vascular diseases are among the leading causes of mortality in the world. Although most people know that the heart must be properly cared for, heart diseases had risen steadily over the last century and has become the leading cause of death for people in the United States. It is predicted that the number of annual deaths will rise to 25 million people by 2030. Medical diagnosis is an important yet complicated task that needs to be executed accurately and efficiently. Heart disease classification systems can help detect anomalies in the patient’s readings and warn the doctors. Association rule mining and classification are two major techniques of data mining.
What is Association rule mining?
Association rule mining is an unsupervised learning method for discovering interesting patterns and their association in large data bases. Whereas classification is a supervised learning method used to find class label for unknown sample. A lot of focus is been devoted towards classification of the heart condition but a very little research is been done towards identifying the risk factors of individual patients and finding the association between the parameters. This research is aimed as improving the classification accuracy while analyzing the associated risk factors.
Literature Research
S. Seema et al. [1] Focused primarily on methods that can find chronic disease by using Naïve Bayes, Decision Tree, Support Vector Machine (SVM) and Artificial Neural Network (ANN) to extract the data stored in previous health records. In order to find the proper output at a Specific scale, a comparative analysis is performed on classifiers.
Jayami Patel et al. [2] Recommend prediction of heart disease using data mining algorithms and machine learning. The main thing is to identify secret patterns through the application of data mining methods. The J48 algorithms dependent on UCI data have the maximum accuracy rate as opposed to LMT.
P. Sai Chandrasekhar Reddy et al. [3] suggested prediction of heart disease using data mining algorithm ANN. Because of the increased cost of diagnosis new technology needed to be built that can predict easily accessible and inexpensive heart diseases, prediction method is used after estimation to know the patient’s condition based on various constraints such as pulse rate, blood pressure, cholesterol etc. The framework is known to be accurate in java.
Ashwini shetty et al. [4] Proposed that a prediction system be built that will diagnose heart disease using the patient’s medical record, and of input attributes were also taken into consideration in developing the system. Data cleaning and data integration were executed after review of data from the dataset.
References:
[1] Dr.S. SeemaShedole, Kumari Deepika, ‘‘Predictive analytics to prevent and
control chronic disease”, https://www.researchgate.net/punlication/
316530782, January 2016.
[2] Jaymin Patel, Prof.TejalUpadhyay, Dr.Samir Patel, “Heart Disease Prediction using Machine Learning and Data Mining Technique”, International Journal of Computer Science and Communication, September 2015-March 2016, pp.129-137.
[3] Mr.P.Sai Chandrasekhar Reddy, Mr.Puneet Palagi, S.Jaya, ‘‘Heart disease
prediction using ANN algorithm in data mining”, Int. J. Comput. ‘‘Heart
disease Prediction using Machine Learning and Data mining Technique”,
March 2017.
[4] A. Ashwini Shetty, Chandra Naik, Different data mining approaches for
predicting heart disease, Int. J. Innov. Sci. Eng. Technol. 5 (2016) 277281.
Data Set
Data Set
The heart disease data sets considered in the study are obtained from a UCI machine learning benchmark repository and IEEE database.
https://archive.ics.uci.edu/ml/datasets/heart+disease
https://ieee-dataport.org/open-access/heart-disease-dataset-comprehensive
About the attributes:
The following are the parameters depicted in the dataset. In order to find out the role of each parameter in causing heart attack, we first identify the parameters which are abnormal. Then, the classification accuracy of each individual parameter is calculated. More the accuracy, more the role of the parameter in causing the heart attack. Then the rules of association are applied on the data to find out the combination of parameters which are linked with one another and thereby a find a combination of features that are more responsible for causing heat attacks.
age: age in years
sex: sex (1 = male; 0 = female)
cp
Value 1: typical angina
Value 2: atypical angina
Value 3: non-anginal pain
Value 4: asymptomatic
considering 1,2,4 as danger, 3 is normal
trestbps: resting blood pressure (in mm Hg on admission to the hospital)
>120 abnormal
chol
>200 abnormal
Fbs: (fasting blood sugar > 120 mg/dl) (1 = true; 0 = false)
restecg: resting electrocardiographic results
Value 0: normal
Value 1: having ST-T wave abnormality (T wave inversions and/or ST elevation or depression of > 0.05 mV)
Value 2: showing probable or definite left ventricular hypertrophy by Estes' criteria
Thalach: maximum heart rate achieved
>100 abnormal
Exang: exercise induced angina (1 = yes; 0 = no)
oldpeak = ST depression induced by exercise relative to rest
slope: the slope of the peak exercise ST segment
-- Value 1: upsloping
-- Value 2: flat
-- Value 3: downsloping
Ca: number of major vessels (0-3) colored by fluoroscopy
thal: 3 = normal; 6 = fixed defect; 7 = reversible defect
Target Variable: diagnosis of heart disease (angiographic disease status)
-- Value 0: < 50% diameter narrowing
-- Value 1: > 50% diameter narrowing
Rule of association:
https://towardsdatascience.com/association-rules-2-aa9a77241654
Columns Type
age object
sex object
cp object
trestbps object
chol object
fbs object
restecg object
thalach object
exang object
oldpeak object
slope object
ca object
thal object
target object
dtyp object
Hypothesis / Research Question(s)
Hypothesis / Research Question(s)
1. What is the risk factor associated with different parameters?
2. How does the risk factor vary from person to person?
3. What is the association between the parameters?
4. How does a combination of parameters effect the risk factor for a person?
5. How accurately can the model classify the data?
Implementation (Model)
Implementation (Model)
Classifiers under study:
Random Forest: Random forest is a flexible, easy to use machine learning algorithm that produces, even without hyper-parameter tuning, a great result most of the time. It is also one of the most used algorithms, because of its simplicity and diversity (it can be used for both classification and regression tasks). In this post we'll learn how the random forest algorithm works, how it differs from other algorithms and how to use it.
Decision Tree Algorithm:
Decision Tree algorithm belongs to the family of supervised learning algorithms. Unlike other supervised learning algorithms, the decision tree algorithm can be used for solving regression and classification problems too.
Logistic regression: Logistic regression is one of the most popular Machine Learning algorithms, which comes under the Supervised Learning technique. It is used for predicting the categorical dependent variable using a given set of independent variables.
SVM: A support vector machine is a very important and versatile machine learning algorithm, it is capable of doing linear and nonlinear classification, regression and outlier detection. Support vector machines also known as SVM is another algorithm widely used by machine learning people for both classification as well as regression problems but is widely used for classification tasks. It is preferred over other classification algorithms because it uses less computation and gives notable accuracy. It is good because it gives reliable results even if there is less data.
Youtube Link Phase 1: https://youtu.be/SdTAVwRCxuc
Youtube Link Phase 1: https://youtu.be/SdTAVwRCxuc
PHASE II: EDA AND MODEL CONSTRUCTION
Data Cleansing:
Data cleansing is the process of identifying and resolving corrupt, inaccurate, or irrelevant data. This critical stage of data processing — also referred to as data scrubbing or data cleaning — boosts the consistency, reliability, and value of your company’s data.
Common inaccuracies in data include missing values, misplaced entries, and typographical errors. In some cases, data cleansing requires certain values to be filled in or corrected, while in other instances, the values will need to be removed altogether.
Also some of the things that has to be made sure before start of the analysis is to make sure the names of the attributes are correct and also to check for missing values in order to make sure the results or the analysis is not manipulated.
Data Preparation:
Data preparation is the process of cleaning and converting raw data before processing and analysis. It's an important step before the process that often includes reformatting data, performing data corrections, and combining data sets to improve data. Data preparation is often a time-consuming process for data practitioners or business users. Still, it is needed as a requirement for putting data in context to transform it into insights and remove bias caused by poor data quality.
checked for missing values and found there are none for the selected dataset. And the dataset is labelled with no any empty or null entries.
As all the records were combined into a single a column separated by comma, I had to use delimiter to divide the values of the 13 attributes.
Check for spacings or errors in the values that contains decimals.
Data Visualizations:
The figure displays that The number of heart conditions in Male are considerably higher than female. As can be seen from the bar graph that Male are much prone to heart diseases or conditions than the females.
The plot shows that the heart conditions increased drastically in the age groups of 55-63. In detail, people between the age 55 to 63 are more prone to heart diseases. People with ages 57, 58, 59 has the highest count of heart disease.
From this plot, it can be seen that More heart conditions occurred in patients with Chol > 230. Which means people with Cholesterol levels greater than 230 has high chances of getting heart diseases when compared to people with cholesterol levels less than 230.
From this bar plot it can be seen that resting blood pressure was recorded abnormally for more number of patients as 120,130,140,150. So to be clear, at these given heart rates, patient's resting blood pressure was abnormal as the chances of getting a heart disease are higher at these resting rates.
In this bar plot it can be seen that Heart rate in patients varied mainly from 70-144. Which means patients heart rate has high variation which is from 70 to 144 for each target class.
Youtube Link Phase 2: https://youtu.be/JCfTwrxlrJI
Youtube Link Phase 2: https://youtu.be/JCfTwrxlrJI
PHASE III: Model Execution And Interpretation
Building Machine Learning Models:
I have used four machine learning algorithms for my analysis.
The Models are:
1). SVM
2). Logistic Regression
3). Decision Tree
4). Random Forest
1). Support Vector Machine (SVM)
1). Support Vector Machine (SVM)
The Overall Accuracy for the test data is 80.32%
The figure shows the confusion matrix for the support vector machine model. A total of 61 data samples have been given for testing. Out of which 49 samples have been correctly identified. The values of TP, TN, FP, FN are 32, 17, 3 and 9 respectively. The overall accuracy obtained is 80.32.
2).Logistic Regression
2).Logistic Regression
The Overall Accuracy for the test data is 80.32%
The figure shows the confusion matrix for the Logistic Regression model. A total of 61 data samples have been given for testing. Out of which 49 samples have been correctly identified. The values of TP, TN, FP, FN are 32, 17, 3 and 9 respectively. The overall accuracy obtained is 80.32.
3). Decision Tree
3). Decision Tree
The Overall Accuracy for the test data is 80.32%
The figure shows the confusion matrix for the Decision Tree model. A total of 61 data samples have been given for testing. Out of which 49 samples have been correctly identified. The values of TP, TN, FP, FN are 29, 18, 6 and 8 respectively. The overall accuracy obtained is 80.32.
4). Random Forest Regression
4). Random Forest Regression
The Overall Accuracy for the test data is 75.40%
The figure shows the confusion matrix for the Random Forest model. A total of 61 data samples have been given for testing. Out of which 49 samples have been correctly identified. The values of TP, TN, FP, FN are 28, 17, 7 and 9 respectively. The overall accuracy obtained is 75.40.
Association Rule Mining:
Association rule mining is a procedure which is meant to find frequent patterns, correlations, associations, or causal structures from data sets found in various kinds of databases such as relational databases, transactional databases, and other forms of data repositories. And also Association Rule Mining is a Data Mining technique that finds patterns in data. The patterns found by Association Rule Mining represent relationships between items.
In My analysis, Association rules can be applied to extract the relationship in the data attributes. Here the parameters responsible for heart attack are studied in detail. The risk factor analysis on the attributed can be performed with support and confidence values. Support describes the probability that both the body and the head of the rule are in of a transaction as measured by all transactions. The confidence of a rule is defined as the proportion of transactions that Rule body and rule header contain the number of transactions that meet the rulebook. The experimental results show that the thalach is the parameter with highest association will other parameters in the abnormal range. The second and third highest parameters are found out as CP, Chol respectively.
Association Rule Mining Between the parameters:
Association Rule Mining Between the parameters:
Abnormal Association between cp and other parameters:
Support
The abnormal association between cp and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the cp with other parameters testbps, chol, fbs, resecg, thalach, exang, ca and thal. The highest association is found with thalach, chol and testbps in order. The least association is found with fbs. For patients with abnormal cp, the other parameters thalach, chol and testbps are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between trestbps and other parameters:
Support
The abnormal association between trestbps and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the trestbps with other parameters cp, chol, fbs, resecg, thalach, exang, ca and thal. The highest association is found with thalach, chol and testbps in order. The least association is found with fbs. For patients with abnormal trestbps, the other parameters thalach, chol and cp are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between Chol and other parameters:
Support
The abnormal association between chol and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the chol with other parameters testbps, cp, fbs, restecg, thalach, exang, ca and thal. The highest association is found with thalach, testbps and cp in order. The least association is found with fbs. For patients with abnormal chol, the other parameters thalach, cp and testbps are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between fbs and other parameters:
Support
The abnormal association between fbs and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the fbs with other parameters testbps, chol, cp, resecg, thalach, exang, ca and thal. The highest association is found with thalach, testbps and chol in order. The least association is found with restecg. For patients with abnormal fbs, the other parameters thalach, chol and testbps are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between restecg and other parameters:
Support
The abnormal association between restecg and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the restecg with other parameters testbps, chol, fbs, cp, thalach, exang, ca and thal. The highest association is found with thalach, cp, chol and testbps in order. The least association is found with fbs. For patients with abnormal restecg, other parameters thalach, chol and cp are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between ca and other parameters:
Support
The abnormal association between ca and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the ca with other parameters, testbps, chol, fbs, resecg, thalach, exang, cp and thal. The highest association is found with thalach, chol and cp in order. The least association is found with fbs. For patients with abnormal ca, the other parameters thalach, chol and testbps are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between thalach and other parameters:
Support
The abnormal association between thalach and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the thalach with other parameters testbps, chol, fbs, resecg, cp, exang, ca and thal. The highest association is found with cp, chol and testbps in order. The least association is found with fbs. For patients with abnormal thalach, the other parameters cp, chol and testbps are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between exang and other parameters:
Support
The abnormal association between exang and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the exang with other parameters testbps, chol, fbs, resecg, thalach, cp, ca and thal. The highest association is found with thalach, cp and chol in order. The least association is found with fbs. For patients with abnormal exang, the other parameters thalach, chol and cp are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Abnormal Association between thal and other parameters:
Support
The abnormal association between thal and other parameters is presented in the figure. The graph depicts the confidence value associated with the abnormal values of the thal with other parameters testbps, chol, fbs, resecg, thalach, exang, ca and cp. The highest association is found with thalach, cp and chol in order. The least association is found with fbs. For patients with abnormal thal, the other parameters thalach, chol, cp and testbps are also most abnormal. The patients with abnormal ca, thalach, testbps and chol have a very high risk of getting heart attack.
Confidence
Conclusion:
Conclusion:
Heart disease prediction is a challenging task that is under research from many decades. There are several factors that cause heart attacks in patients, so these factors can be used to analyze and predict if a patient is having a risk of getting heart attack. Also the Machine learning models that were built will help predict if a patient is going to have a heart attack. The association analysis helps identify the most prominent risk factor in the patients that can cause heart attack. The analysis of the association can help doctors personalize the treatment based on the patient condition.
Links (Optional)
Links (Optional)
Your Linkedin Profile
Your GitHub Repo: https://github.com/mrudulaj97/Heart-Diseases
Page updated
Report abuse