Multi-Layer Perceptron Model for Dota 2 Game Results from UCI Using MLP Classifier

Abstract

Dota 2 is a genre game Moba in the PC (Personal Computer) system battle arena game online (online) with multiplayer ( bringing together 2 players in 1 machine ). Game Dota 2 consists of 2 opposing teams To get the victory, every team has 5 players who can choose hero 1 from 121 different heroes. Study This discusses the use of the Multi-Layer Perceptron (MLP) model to predict the results Dota 2 game. The author uses the UCI dataset containing historical data of Dota 2 matches, processed and trained with the MLP model using MLPClassifier from the scikit learn Python library. The data preprocessing process includes normalization features and handling of missing data. Training involves hyperparameter selection and validation cross To prevent overfitting. Although the MLP model is successful in predicting results with accuracy high, the author takes notes room For improvement, like additional features or the use of more models complex. In research, This obtained results with Accuracy Train results: 68.06%, Accuracy Test: 58.00%, Accuracy Precision: 58.53%, Accuracy Recall: 73.50%, Accuracy f1: 63.39%.