The paper was accepted in the flagship conference "TENCON 2021", hosted in Auckland, New Zealand. This work is a part of my internship program at IIIT Guwahati.
The link to the paper is: https://ieeexplore.ieee.org/abstract/document/9707283
Download the dataset from this link: https://drive.google.com/file/d/1CRmKbr6EUZ5WY5RJWVh0MjBEuA-aVMDB/view?usp=sharing
Abstract: Asthma is a chronic disease which is affecting a huge population around the world. In this disease, air passages of lungs in a human body become narrow due to inflammation and tightening of the muscles. It causes repeated episodes of wheezing, breathlessness, sleep disturbances, chest tightness, nighttime or early morning coughing etc. Though an occurrence of any one of these symptoms, at a time, cannot be concluded as asthma, but repeated occurrence of the combined symptoms may be concluded as asthma. Therefore, it is highly required to detect asthma as early as possible before getting into the corresponding exacerbation. Diagnosis of asthma in children is a very difficult process. Certain devices may be created that could monitor these symptoms in child, wherein a machine learning model can be deployed to detect the initial development of asthma. We develop one model that could detect early stage of asthma in children based on asthma status of the parents and some of the combination of core symptoms. We used the dataset prepared in phase two of International Study of Asthma and Allergies in Children (ISAAC). We have tried four relevant machine leaning models to select the model with best accuracy. Four models that we have tried are Decision Tree Classifier, Random Forest Classifier, k-Nearest Neighbor and Artificial Neural Network. We finally selected the best model,i.e., Artificial Neural Network with the training accuracy as 95% and test accuracy as 91.6%, respectively.