Prediksi Harga Sepeda Motor Menggunakan Metode Weighted Evolving Fuzzy Neural Network (Wefunn)

ABSTRAK

Harga sepeda motor selalu berubah setiap tahunnya, oleh karena itu diperlukan sebuah
pendekatan dalam memprediksi besarnya harga sepeda motor dengan keakuratan
maksimum. Salah satu dari jenis prediksi kuantitatif adalah prediksi data time series
yakni suatu teknik prediksi yang dibangun menggunakan data runtun waktu pada
periode tertentu. Dalam tugas akhir ini digunakan metode Weighted Evolving Fuzzy
Neural Network (WEFuNN) untuk memprediksi harga sepeda motor berdasarkan data

runtun waktu. WEFuNN merupakan pengembagan dari metode Evolving Fuzzy
Neural Network (EFuNN) yang memiliki struktur hybrid dari metode Fuzzy Inference
System (FIS) dan jaringan saraf tiruan (Neural Network) dengan menerapkan prinsip
Evolving Conection System (ECOS) didalam jaringan. Tingkat keakuratan hasil

prediksi diukur dengan nilai MAPE (Mean Absolute Percentage Error ). Hasil prediksi
WEFuNN didapat hasil error rata-rata (MAPE) yaitu sebesar 1,269% dengan
menggunakan data penjualan real periode Januari 2014 sampai dengan Juli 2014.

Kata kunci: weighted evolving fuzzy neural networks , fuzzy, evolving connectionist
system, fuzzy inference system, peramalan


Universitas Sumatera Utara

WEFUNN METHOD IN FORECASTING THE PRICE OF MOTORCYCLE
SALES

ABSTRACT

Motorcycles price is always changing every year, therefore we need an approach to
predict the magnitude of the price of a motorcycle with a maximum of accuracy. One
of the types of quantitative prediction is forecasting of time series data that is a
prediction technique which is constructed using time series data over a given period.
In this thesis used Weighted Evolving Fuzzy Neural Network (WEFuNN) to predict
the price of a motorcycle based on time series data. WEFuNN is developing a method
of Evolving Fuzzy Neural Network (EFuNN) which has a hybrid structure of the
Fuzzy Inference System (FIS) and artificial neural networks (Neural Network) by
applying the principle of Conection Evolving System (ECOS) in the network. The
level of accuracy of the prediction is measured by the value of MAPE (Mean Absolute
Percentage Error). Results obtained WEFuNN prediction average error (MAPE) is
equal to 1,269% by using real sales data for the period of January 2014 through July
2014.


Keywords : weighted evolving fuzzy neural networks, fuzzy, evolving connectionist
system, fuzzy inference system, forecasting

Universitas Sumatera Utara