UTAR Institutional Repository

Esport valorant: Optimizing hyperparameters of win rate prediction model using derivative-free method

Chang, Yang (2024) Esport valorant: Optimizing hyperparameters of win rate prediction model using derivative-free method. Final Year Project, UTAR.

[img]
Preview
PDF
Download (4Mb) | Preview

    Abstract

    This research investigates the application of machine learning to predict outcomes in Valorant, a rapidly-growing first-person shooter game within the esports field. This project developed two predictive models, neural network and XGBoost, and optimized using Bayesian optimization and random search to optimize their hyperparameters. The findings shows that the efficiency and capability of Bayesian optimization over random search in terms of both model performance and computational efficiency. To enhance public accessibility and usability, this project has created web APIs and a user-friendly graphical interface. This research contributes significantly to the field of esports analytics and provides practical tools for predicting Valorant gameplay outcomes. By offering a robust and efficient predictive system, this research aims to support informed decision-making, enhance the overall experience for pro-players and enthusiasts of this popular game, and potentially suggest strategic development within the Valorant community. Furthermore, the findings of this project may serve as a valuable reference for future research exploring the application of machine learning to other esports games. Specifically, this research provides insights into the effectiveness of Bayesian optimization in optimizing machine learning models for complex tasks. By comparing Bayesian optimization to random search, it highlights the benefits of Bayesian optimization's ability to intelligently balance between explore and exploit the hyperparameter search space, leading to more efficient and effective model optimization. Additionally, our study noted that the importance of developing user-friendly interfaces and APIs to make win rate prediction tools accessible to a wider audience, fostering collaboration and innovation within the esports community.

    Item Type: Final Year Project / Dissertation / Thesis (Final Year Project)
    Subjects: Q Science > Q Science (General)
    T Technology > T Technology (General)
    Divisions: Faculty of Information and Communication Technology > Bachelor of Computer Science (Honours)
    Depositing User: ML Main Library
    Date Deposited: 27 Feb 2025 14:55
    Last Modified: 27 Feb 2025 14:55
    URI: http://eprints.utar.edu.my/id/eprint/6945

    Actions (login required)

    View Item