1. Enhancing Serious Game Experience Through In-Game Radio Using Context-Aware Recommender System Based on Player Behavior.
- Author
-
Damastuti, Fardani Annisa, Firmansyah, Kenan, Arif, Yunifa Miftachul, Dutono, Titon, Barakbah, Aliridho, and Hariadi, Mochamad
- Subjects
MACHINE learning ,RECOMMENDER systems ,REAL-time computing ,FINITE state machines ,REINFORCEMENT learning - Abstract
This article describes the creation and implementation of an in-game radio context-aware recommender system (CARS). To increase player engagement, music is selected based on time of day, in-game seasons, weather, and player conditions. The main purpose is to customize in-game radio features, which often don't respond to changing gameplay. Real-time weather (sunny, rainy, stormy), player circumstances (boat health state, in-game currency, player activity status), time of day (morning, afternoon, evening, night), and in-game seasons (sunny, damp) were collected. Processing the data yielded machine learning model characteristics. The system used deep learning and hybrid filtering to recognize complex contextual data and player behavior patterns. To test the recommender system, precision, recall, accuracy, and F1 score were used. Compared to collaborative filtering, content-based filtering, multi-criteria recommender systems, ANN-based finite state machines, and hybrid optimization methods, our system achieved an F1 score of 0.64-0.71 and an accuracy of 0.68-0.75 in various settings. This beats older approaches, which have 0.60-0.65 accuracy under inactive conditions. An internal iterative testing dataset of real-time player interactions and ingame environmental circumstances was used to train and assess our models. F1 scores for "Healthy Boat High Money" and "Sunny Morning" were 0.71 and 0.70, respectively, while accuracy was 0.75 and 0.74. These findings demonstrate the potential of real-time contextual data in recommender systems to boost user engagement. The essay provides a robust system architecture and powerful machine learning models for real-time data processing and low-latency suggestions. Expanding the dataset, exploring new domains, and using reinforcement learning can show the system's flexibility and performance. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF