Developing a Prediction Model for Coronavirus-2019 infections Base on Patient Symptoms

Abstract

The objectives of this research are: 1) to develop a prediction model for coronavirus-2019 infection from patient symptoms using data mining techniques 2) to compare and evaluate the performance of the model. The patient condition dataset was collected according to the guidelines of the World Health Organization and the All India Institute of Medical Sciences (AIIMS). This dataset contains a total of 5,434 records, which are divided into 4,348 records for learning and 1,086 records for testing. 10 significant features were carefully selected from a list of 21 to be used in the data mining algorithm's process of learning.  The CRIPS-DM approach served as the main research methodology for this study. The learning methodology consists of 8 algorithms: Random Forest, Naïve Bayes, Support Vector Machine, Neural Network, Ada-Boot, K-Nearest Neighbor, Logistic Regression, and Decision Trees. Learning processes and prediction models' performance is evaluated through classification accuracy, precision, sensitivity, and specificity. The study's findings showed that neural network algorithms provide the best learning performance. When various parameters were adjusted in this approach, the prediction model achieved the highest efficiency of 91.40% with the learning rate set to 0.1, ReLU as the activation function, and the maximum number of iterations set to 200. To summarize the results were that, the development of a prediction model for coronavirus-2019 infection based on patient symptoms using a neural network algorithm has shown best predictive performance.