Plant Growth Phase Classification Using Deep Neural Network (Case Study of ASF in Poso District, Central Sulawesi Province)

An innovation developed through a combination of satellite data with official data to provide a solution to the limitations of the Area Sample Framework (ASF) survey where surveyors have to go directly to places that are sometimes difficult to reach and require a relatively long time, the Central Statistics Agency (BPS) suggested using Landsat-8 satellite imagery with Deep Neural Network Method (DNN) to classify rice plant growth phases. Data from Landsat-8 which has the characteristics to see land cover, especially plants. Apart from the band, the variables in this study were added to the vegetation index calculated from satellite data and combined with official data. One of the classification methods used is Deep Neural Network. This study aims to compare the methods between Artificial Neural Network (ANN) and DNN in classifying rice growth phases and predicting rice growth phases using DNN. With split data stratified 5-fold cross validation and data normalization using a robust scaler, the classification results show the average performance in terms of accuracy, precision, sensitivity, f1-score, Cohen Kappa index and Average Precision (AP) values. Based on several performance evaluations of the two methods from both ANN and DNN there is no significant difference.