COMPARISON OF ANN METHOD AND LOGISTIC REGRESSION METHOD ON SINGLE NUCLEOTIDE POLYMORPHISM GENETIC DATA

Abstract

This study aims to determine the goodness of classification using the ANN method on Asthma genetic data in the R program package, namely SNPassoc. SNP genetic data was transformed using codominant genetic traits, namely for genetic data AA, AC, CC were given a score of 0, 0.5 and 1, respectively, while CC, CT and TT were scored 0, 0.5 and 1, respectively. The scoring is based on the smallest alphabetical order given a low score. The average accuracy, precision, recall and F1 score were determined using the neural network method if the genetic code was used with variations in the proportion of test data 10%, 20%, 30% and 40% and repeated B = 1000 times. The results obtained were compared with the logistic regression method. If 20% test data is used and the ANN method is used, the accuracy, precision, recall and F1 scores are 0.7756, 0.7844, 0.9844 and 0.8728, respectively. When all information from various countries is used in the Asthma genetic data, the logistic regression method gives higher average accuracy, precision and F1 scores than the ANN method, but the average recall is the opposite. When a separate analysis is performed for each country, the logistic regression method gives higher accuracy, precision, recall and F1 scores in the ANN method compared to the logistic regression method.