Palm Oil Prediction Production Using Extreme Learning Machine

Abstract

The total production of Indonesian palm oil (CPO) in 2018 reached 43.9 million tons, with a land area of 12.3 million hectares.However, every month there are still many companies that have problems in predicting palm oil production. Problems in predicting thisproduction can be solved by calculation methods in the field of artificial neural networks, namely the Extreme Learning Machine (ELM)method. This method can solve linear and non-linear data problems and provide better average computation compared to other methods inpredicting oil palm production. The data used is palm oil production data at PT Indo Palm Oil Labuhan Batu with a total of 297 in the period2017-2018. While the parameters used are planting age, land area, number of trees, and yields. The results of the best-hidden neuron testare 13 with 2 technical data features and the training data pattern is pattern 1. The average MAPE value is 20.1% with the fastestcomputing time is the use of the number of hidden neurons 2. So based on the test results, the method ELM has a predictive model withquite good performance because the MAPE value is in the range of 20% -50%.