The comparison between recurrent neural network and grey model to predict Indonesia tuberculosis morbidity rate

D. Harlianto, S. Mardiyati, D. Lestari

Research output: Contribution to journalConference articlepeer-review

Abstract

Many insurance companies offer health insurance products that cover the risk of tuberculosis (TB). The risk of disease is generally stated as the morbidity rate that is the ratio of the number of residents suffers from an illness to the total population. In 2018, the World Health Organization (WHO) reported that Indonesia was at rank two for the case numbers of TB, so Indonesia has a high risk of this disease. The recurrent neural network (RNN) and grey model are two models that can be employed to predict TB morbidity rates. In this research, the accuracy of these two models was compared. The results of this research may give the insurance company an assist to choose an appropriate mathematical model to provide a competitive and profitable premium. TB morbidity rate in a certain year had been predicted based on the past several year's morbidity rates as model input. The size of past years data used as model input was made varied to observe how information availability influences the model accuracy measured by mean squared error (MSE) and mean absolute percentage error (MAPE). The results show that the grey model has better accuracy when the small data used as input. On the other hand, the accuracy of RNN is not affected significantly by the setting of the data input size used in this research.

Original languageEnglish
Article number012093
JournalJournal of Physics: Conference Series
Volume1722
Issue number1
DOIs
Publication statusPublished - 7 Jan 2021
Event10th International Conference and Workshop on High Dimensional Data Analysis, ICW-HDDA 2020 - Sanur-Bali, Indonesia
Duration: 12 Oct 202015 Oct 2020

Fingerprint

Dive into the research topics of 'The comparison between recurrent neural network and grey model to predict Indonesia tuberculosis morbidity rate'. Together they form a unique fingerprint.

Cite this