Prediksi Skor Pertandingan Sepak Bola menggunakan Neuroevolution of Augmenting Topologies dan Backpropagation

Welly Winata (Program Studi Informatika)
Lily Puspa Dewi (Program Studi Informatika)
Alvin Nathaniel Tjondrowiguno (Program Studi Informatika)

Jurnal Infra
Vol 8, No 1 (2020)

Football, or soccer is the most popular sport in the world. Whatmakes football special is the uncertainty and unpredictable result.There are a lot of factors that can affect the result of a footballmatch, such as strategy, skill, or even luck. Therefore, predictingthe outcome of football match can be challenging yet interestingtask.This research started with neuroevolution of augmentingtopologies, which useful to find the structur of a neural network.Then, the network produced by NEAT is optimized usingbackpropagation. Player ratings, team ratings, and playerposition are used as features of neural network.The hightest accuracies achieved are 81.5% on the final resultpredicting, and 48% on score predicting, were obtained throughNEAT network that optimized by backpropagation, with playerratings, team ratings, and total position from each sectors areused as features.However, on real life test, the player and team ratings areunknown. To calculate the player and team ratings, averagesmethods are used. Unfortunately, the network performed poorlycausing the accuracies to dropped significantly. Lack ofconsistency from player ratings are believed to be the mainproblem on calculating the player and team ratings.