This master's thesis explores the profiling of video gamers based on data obtained exclusively from the controllers they use during gameplay. The study focuses on four popular games: Battlefield V, PUBG, Elden Ring, and Dark Souls 3, representing two distinct genres: first-person shooters (FPS) and role-playing games (RPG). The objective is to investigate the viability of using data gathered from two different controllers, the Nintendo Switch Pro Controller and the DualShock 4, to train an AI model utilizing machine learning algorithms. The experimental procedure involved participants playing each game for a duration of 15 minutes using each controller, resulting in a total of 30 minutes of gameplay per game. Throughout the gameplay sessions, data from various input elements such as buttons, triggers, and sensors was collected to construct the training dataset for the AI model. The central objective of this research is to classify video gamers in six specific scenarios. These scenarios involve training and testing the AI model using various combinations of games and controllers. Initially, the experiments are conducted by training and testing on a single game. Subsequently, the process is repeated with training on one game and testing on another game within the same genre. And finally, the experiments involve training on one game and testing on another game from a different genre. In all cases, the experiments are executed first using one controller for both training and testing, and then by utilizing one controller for training and another for testing. We conducted the experiments with 23 participants, collecting almost 50 hours worth of data. The highest results were obtained in the most correlated scenario which reached an F1-score as high as 87%. The outcomes have implications for personalized gaming experiences, adaptive difficulty adjustments, and targeted game recommendations. Furthermore, the study will provide insights into the effectiveness of employing controller data to differentiate player behaviors and preferences within specific genres as well as across different genres.
This master's thesis explores the profiling of video gamers based on data obtained exclusively from the controllers they use during gameplay. The study focuses on four popular games: Battlefield V, PUBG, Elden Ring, and Dark Souls 3, representing two distinct genres: first-person shooters (FPS) and role-playing games (RPG). The objective is to investigate the viability of using data gathered from two different controllers, the Nintendo Switch Pro Controller and the DualShock 4, to train an AI model utilizing machine learning algorithms. The experimental procedure involved participants playing each game for a duration of 15 minutes using each controller, resulting in a total of 30 minutes of gameplay per game. Throughout the gameplay sessions, data from various input elements such as buttons, triggers, and sensors was collected to construct the training dataset for the AI model. The central objective of this research is to classify video gamers in six specific scenarios. These scenarios involve training and testing the AI model using various combinations of games and controllers. Initially, the experiments are conducted by training and testing on a single game. Subsequently, the process is repeated with training on one game and testing on another game within the same genre. And finally, the experiments involve training on one game and testing on another game from a different genre. In all cases, the experiments are executed first using one controller for both training and testing, and then by utilizing one controller for training and another for testing. We conducted the experiments with 23 participants, collecting almost 50 hours worth of data. The highest results were obtained in the most correlated scenario which reached an F1-score as high as 87%. The outcomes have implications for personalized gaming experiences, adaptive difficulty adjustments, and targeted game recommendations. Furthermore, the study will provide insights into the effectiveness of employing controller data to differentiate player behaviors and preferences within specific genres as well as across different genres.
Video Gamers Profiling Based on Controller Usage and Behavioral Data
ZIN, RICCARDO
2022/2023
Abstract
This master's thesis explores the profiling of video gamers based on data obtained exclusively from the controllers they use during gameplay. The study focuses on four popular games: Battlefield V, PUBG, Elden Ring, and Dark Souls 3, representing two distinct genres: first-person shooters (FPS) and role-playing games (RPG). The objective is to investigate the viability of using data gathered from two different controllers, the Nintendo Switch Pro Controller and the DualShock 4, to train an AI model utilizing machine learning algorithms. The experimental procedure involved participants playing each game for a duration of 15 minutes using each controller, resulting in a total of 30 minutes of gameplay per game. Throughout the gameplay sessions, data from various input elements such as buttons, triggers, and sensors was collected to construct the training dataset for the AI model. The central objective of this research is to classify video gamers in six specific scenarios. These scenarios involve training and testing the AI model using various combinations of games and controllers. Initially, the experiments are conducted by training and testing on a single game. Subsequently, the process is repeated with training on one game and testing on another game within the same genre. And finally, the experiments involve training on one game and testing on another game from a different genre. In all cases, the experiments are executed first using one controller for both training and testing, and then by utilizing one controller for training and another for testing. We conducted the experiments with 23 participants, collecting almost 50 hours worth of data. The highest results were obtained in the most correlated scenario which reached an F1-score as high as 87%. The outcomes have implications for personalized gaming experiences, adaptive difficulty adjustments, and targeted game recommendations. Furthermore, the study will provide insights into the effectiveness of employing controller data to differentiate player behaviors and preferences within specific genres as well as across different genres.File | Dimensione | Formato | |
---|---|---|---|
Zin_Riccardo.pdf
accesso aperto
Dimensione
2.64 MB
Formato
Adobe PDF
|
2.64 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/52259