Classification of COVID-19 Data Using Decision Tree Optimized with Particle Swarm Optimization and Genetic Algorithm

To date, COVID-19 is still spreading in many countries and is endemic. To support medical personnel in diagnosing patient conditions, early detection can be done with a machine learning approach based on historical data of previous COVID-19 sufferers. This study applies the C4.5 algorithm with hold-out and k-fold cross validation techniques to classify the status of patients exposed and not exposed to COVID-19 recorded in COVID-19 data for the world and a city in Indonesia. Dimensionality reduction with principal component analysis (PCA) was also applied to the global COVID-19 data to reduce the features in the data. To improve the classification ability of C4.5, two optimization algorithms which are particle swarm optimization and genetic algorithm are applied with a hybrid approach. The best modelling results on city data were obtained using PSO with an increase in accuracy by 8.27%, precision by 24.49%, and recall by 3.75%. The best optimization results on world data were obtained using GA with an increase in accuracy of 0.31%, precision of 0.25%, and recall of 0.07%. The results showed that the application of PSO and GA succeeded in improving the performance of the C4.5 algorithm on both data even though the world data did not experience a significant increase.