loading page

Analysing and Evaluating the Performance of Multi-Layer Perceptrons and Consequently Intelligent Algorithms in Highly Chaotic and Highly Dimensional Environments Through Fantasy Football
  • Arnav Shah
Arnav Shah
N/A, Dhirubhai Ambani International School

Corresponding Author:[email protected]

Author Profile

Abstract

With the passage of time, the ability of intelligent software in making predictions of seemingly random, chaotic events increases. Most real-world uses of multi-layer perceptrons are in highly chaotic environments. Such environments include the prediction of economic trends and weather phenomena to the modelling of human behaviour. This study aims to analyse the power of Multi-Layer Perceptrons (MLPs) in predicting the outcome of sporting events, in particular, a football (soccer) game. Any sporting event is dominated by chaos, where small changes in the environment can drastically affect the outcome. This study does the same by predicting, based on multi-dimensional data, the point return of English Premier League (EPL) players in the official Fantasy Premier League forum (FPL). The study uses MLPs as a means to optimise FPL team selection and thus predicts an average point return for the team through the summation of point average predictions for each player individually. This is then compared with the actual FPL gameweek return of the selected team and evaluated. The study found that an MLP can predict the return of each player with an RMSE of approximately 2 points. The study shows the extent of the capability of MLPs in finding ordered patterns in chaotic events.