Date of Publication :8th August 2024

Abstract:For several reasons, including team and player growth as well as coach and sports expert decision-making, sports prediction is vital. The main objective of this article is to apply machine learning (ML) techniques to build a data-driven model that can forecast the results of NBA league games. The work done in this article starts with data processing and continues with model construction using five different ML models including Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF) and Artificial Neural Networks(ANN). These five models were then evaluated using different metrices such as accuracy, precision, recall and F1-score. This study's systematic approach offers a flexible framework that may be used in a range of sports analytics contexts. Based on past extended index averages, the model projects symmetric extended indices for both teams playing in upcoming games. Testing and training sets from multiple seasons are used to evaluate the suggested model. The variables pertaining to teams, players, and opponents are the main emphasis of this research. These variables include field goal attempts, three-point attempts, made free throws, attempted free throws, and offensive rebounds, among others. The algorithms performed differently, as demonstrated by the results, with the RF Algorithm coming out on top with 84% accuracy. This shows how well RF predicts the likelihood of a winning team, outperforming SVM (74%), LR (82%), ANN (82%), KNN (79%). The proposed is intended to help the coaches to select the right set of team members to improve the chance of winning in the tournament. This work is still limited in terms of data quality and model restrictions. Nonetheless, sports professionals looking for practical insights should immediately consider the ramifications of the findings. This work points to future directions for research, encouraging the creation of more intricate algorithms, detailed feature analysis, and the integration of temporal patterns for comprehensive predictive accuracy.

Reference :