Penerapan Data Mining Untuk Memprediksi Jumlah Produksi Kelapa Sawit Menggunakan Metode Regresi Linear Berganda (Studi Kasus: PT. Padasa Enam Utama)

Saragih, Loisa (2021) Penerapan Data Mining Untuk Memprediksi Jumlah Produksi Kelapa Sawit Menggunakan Metode Regresi Linear Berganda (Studi Kasus: PT. Padasa Enam Utama). Other thesis, Universitas Islam Riau.

Text 163510400.pdf Download (5MB)

Abstract

Palm oil (Elaeis guineensis Jacq) is one of the plantation sectors that has benefits in improving the economy in society, because this plantation commodity has a very high level of production profitability. In general, the amount of harvest produced by palm oil can not reach the target, usually due to harvesting. In a company, such as PT. Padasa Enam Utama, the amount of harvested from oil palm production is not automatically known whether the amount of yield from the production is increasing or decreasing, so that the plantation assistant are unable to predict future production yield. This is because the calculation process is still using manual analysis. In this research will use some prediction features that are land area, amount of harvest, fertilizer, amount of harvesters and amount of crops produced. The method used is the Double Linear Regression method. This method has the advantage of structured and easy to understand data decoding as well as accuracy in prediction results. The calculation process starts from the process of pattern determination, training of amount of training data and test data, calculation of predictive error values and generating final value. The data used is production data in 2017-2019 with a total of 235 data, a total of 205 training data and 30 testing data. The result of the research was obtained the future prediction value with the testing phase of data testing. In testing the results of the questionnaire presentase use the likert scale obtained a percentage value of 88.333%. At the test stage the predictive accuracy level results using the Mean Absolute Percentage Error (MAPE) method with a 12.308% error rate percentage resulting in the system being worth using.

Actions (login required)

View Item