Multivariate Data Model Prediction Analysis Using Backpropagation Neural Network Method

Indonesia has a relatively large population, making health an important necessity. Census data has been widely available in accordance with problems in health. With the available data, researchers have a great opportunity to conduct research. Research on predictions in health is nothing new. By making predictions, it can help in making decisions quickly and accurately decisions that can be determined by the health sector or other fields. Prediction methods have been developed, one of which is an artificial neural network (ANN). In the ANN method, there is a Backpropagation Neural Network (BPNN) model. In general, the BPNN method is implemented in one input data model. Therefore, it requires an analytical study that uses a lot of data models to be implemented into the BPNN method. In this study, five data models were used with the number of variables and the amount of data that were different. Processed with the BPNN method and outputs the accuracy and execution time of all data models. From experiments and analyzes conducted on 5 kinds of data models, data model 4 has the number of variables 19 and the number of data 392 yields accuracy of 98.718% and a relatively slow execution time of 27.798 seconds. The highest execution time is found in the data model 3 with the number of variables 13 and the amount of data of 589 yields an accuracy of 95,385% and execution time of 45,442 seconds.