MODEL FEED FORWARD NEURAL NETWORK (FFNN) DENGAN ALGORITMA PARTICLE SWARM SEBAGAI OPTIMASI BOBOT (Studi Kasus : Harga Daging Sapi dari Bank Dunia Periode Januari 2007 – Desember 2018)

*Faisal Fikri Utama  -  Departemen Statistika, FSM, Universitas Diponegoro, Indonesia
Budi Warsito  -  Departemen Statistika, FSM, Universitas Diponegoro, Indonesia
Sugito Sugito  -  Departemen Statistika, FSM, Universitas Diponegoro, Indonesia
Published: 28 Feb 2019.
Open Access Copyright 2020 Jurnal Gaussian
License URL: http://creativecommons.org/licenses/by-nc-sa/4.0

Citation Format:
Abstract

Beef is one of the important food commodities to fulfill the nutritional adequacy of humans. The World Bank notes the beef prices that are exported worldwide every month. For this reason, those data becomes a predictable series for the next period. Feed Forward Neural Network is a non-parametric method that can be used to make predictions from time series data without having to be bound by classical assumptions. The initiated weight will be evaluated by an algorithm that can minimize errors. Particle Swarm Optimization (PSO) is an optimization algorithm based on particle speed to reach the destination. The FFNN model will be combined with PSO to get predictive results that are close to the target. The best architecture on FFNN is obtained with 2 units of input, 1 unit of bias, 3 hidden units, and 1 unit of output by producing MAPE training 11.7735% and MAPE testing 8.14%. According to Lewis (1982) in Moreno et. al (2013), the MAPE value below 10% is highly accurate forecasting.

 

Keywords: Feed Forward Neural Network (FFNN), Particle Swarm Optimization (PSO), neurons, weights, predictions.

Keywords: Feed Forward Neural Network (FFNN), Particle Swarm Optimization (PSO), neurons, weights, predictions.

Article Metrics: