- Demo
- Overview
- Motivation
- Technical Aspect
- Installation
- Code Run
- Technologies Used
- Team
- License
- Credits
link :https://www.kaggle.com/rashikrahmanpritom/heart-attack-analysis-prediction-dataset
This is a simple classification problem trained on python scikit-learn library.The classification model takes the independent variable eg. age,sex,cholestrol,blood pressure etc.,from heart attack data set to predict whether the person will get heart attack or not.
In this project chi-squre test has been used for to checkfeature importance of categorical variableand independent t-test is used to compare the mean of variables grouped on the basis of output category and by finding correlation among numerical variables to get the importance of each variable in deciding output.Features renaming also done with help of google,you can find here.
We also use Pipeline method to apply 10 classifiction algorithms(1.logistic Regression 2.Decisiontreeclassifier 3.Randomforestclassifier 4.GaussianNB 5.KNN 6.Gboost Classifier 7.AdaboostClassifier8.SGDClassifier 9.SVC 10.MLP Classifier) to get the best accuracy which i have got in KNN=82.4%.Then i also apply for loop to get the best random state producing good accuracy and then we have got accuracy of 90% with choosing appropriate n_nieghour parameter which is n_neighour=6. After plotting AUC-ROC curve we have got the value AUC=93% which is a good value for a model.
In india, every year 30000 people get died from heart attack and this count is increasing continuously every year thats why i was very curious to know about the effect of various factors which causes the heart attack ,which factor is more responsible for haert attack and what is corr- relation among those factors.In India most of the people who died from heart attack are the politicial,celebrity,bussinesman and govt.officials.I was very curious to know why mostly these people are died from heart attack, which factor is different in their bodies.
1.Using Pandas and numpy data importing,data cleaning,data preparation is done.
2.Statical test is done with scipy and visualisation with seaborn & matplotlib for feature analysis,EDA and to get the feature importance of variables.
3.trained 10 classification models using Pipeline and scikit learn,choosen one having the best accuracy.
4.Used for loop and gsearchfor hyperparameter tuning and appropriate random state to improve the accuracy.
5.used AUC-ROC curve to check the ability of our model classifier to distinguish between classes.
The code is written in python(version 3.8.5) in jupyter notebook IDE,with Anaconda python distribution. You can download Anaconda here
[
NaveenKumarMaurya/HeartAttackPrediction is licensed under the Apache License 2.0.
A permissive license whose main conditions require preservation of copyright and license notices. Contributors provide an express grant of patent rights. Licensed works, modifications, and larger works may be distributed under different terms and without source code.
1.Kaggle
2.Google
This project wouldn't have been possible without this tool. It saved my enormous amount of time while collecting the data