Multi-Objective Multi-Instance Learning: A New Approach to Machine Learning for eSports.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Traditional eSports prediction models focus on individual players, neglecting crucial team dynamics.
• Team cooperation is vital for success in eSports, necessitating a team-centric prediction approach.

Purpose of the Study:

• To develop a novel machine learning approach for accurate prediction of eSports match outcomes.
• To address the limitations of player-centric models by introducing a team-centric classification method.

Main Methods:

• Proposed Multi-Objective Multi-Instance Learning (MOMIL), a novel approach applying multi-instance learning to eSports win prediction.
• Jointly considered player objectives within a team to capture inter-player relationships during classification.
• Utilized entropy as a measure of impurity for decision tree construction.

Main Results:

• The MOMIL approach demonstrated superior accuracy compared to standard classification techniques.
• Models were built using season-based data, a departure from previous studies.
• Achieved up to 95% accuracy in eSports win prediction, surpassing state-of-the-art methods on the same dataset.

Conclusions:

• MOMIL effectively captures team dynamics for improved eSports win prediction.
• The team-centric, multi-instance learning framework offers a significant advancement in the field.
• This research provides a highly accurate and robust method for predicting eSports match outcomes.