NBA LEAGUE 2019-2020 WINNER PREDICTION USING ARTIFICIAL NEURAL NETWORK

• Main Author: Setia Atmadja, William
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/47776
• Institution: Institut Teknologi Bandung

The National Basketball Association, also known as NBA is a famous annual basketball league based in the United States of America. Predicting or guessing the outcome of an NBA match has become a thing due to the league’s popularity worldwide. Through deep learning, using the previous games statistics of each and every competing teams as an input makes it possible to build a model designed to accurately predict the outcome of those matches. The aforementioned deep learning model is called Multi-layer perceptron model, or in short MLP. MLP is a type of artificial neural networks. In this final project, 2 MLP models will be built and trained. The first MLP model is constructed to predict the outcome of NBA matches in form of binary classification problem. Meanwhile, the second MLP model is trained to solve regression problems, predicting the score ratio of NBA matches on top of their outcome. Many variations of nodes and hidden layers are imbued when devising these MLP models in order to find out the best MLP model architecture to predict the NBA matches. For the binary classification problem, the best MLP model is composed of 3 hidden layers with 20 nodes within. On the other side, MLP model made up of 5 hidden layers with 40 nodes each is the best model to predict the regression problem. Both models predict the same thing; Milwaukee Bucks as the NBA league 2019-2020 champion with a score ratio of 0.527 score ratio during the final.