Інтелектуальна інформаційна технологія аналізу комп'ютерної гри Dota 2

Abstract

This work presents an intelligent information technology for analyzing the computer game Dota 2, combining machine learning methods with automated processing of gameplay data to recommend the optimal hero. Modern analytical platforms were reviewed, and their limitations were identified, highlighting the need for an adaptive system capable of considering draft context and individual player characteristics. A data acquisition, cleaning, and structuring algorithm was developed, along with a transformer-based model. Testing on three datasets of 40 matches demonstrated the effectiveness of the proposed technology: a 60% win rate compared to 52.5% using Dotabuff recommendations and 50% with intuitive hero selection.