A Multi-Level Adaptive Loot Box Recommendation System for Video Games

Document Type : Primary Research paper

Authors

1 Research Scholar, Sathyabama University, Chennai

2 Sr. Consultant, Bassure Solutions Pvt. Ltd, Chennai

Abstract

Loot Box is one of the most important economies in video Games especially in massively multiplayer online (MMO) games. With Free to Play games growing in popularity, loot boxes became a primary monetisation technique. Loot boxes consist of virtual items that are consumed by players. The items vary from a new weapon to a skin. Loot boxes are one of the reward elements that create continuous interest among players to play games. On the other hand, publishers can generate revenues even after the game is released. These advantages have made loot boxes quite popular in the video game industry. Loot boxes have become one of the monetizing methods for video games. An online multiplayer accommodates different types of players and also the design incorporates opportunities for different types of fun. Considering this design philosophy, defining loot boxes with predefined items in it may cause disappointments to players as the item does not fit for them. In addition to causing negative engagement, such scenarios will not favor purchases that affect the game economy. Recommending loot items adaptive to the traits of the players and their play style could be a desired solution. We recommend a novel approach based on machine learning algorithms to solve this problem. The approach couples to Machine learning solutions together to generate loot item suggestions. The system uses Random Multimodal Deep Learning based classification approach to identify player profile and classification. The system identifies the player trait and the play style using sentiment analysis. The model employs Bartle player classification to evaluate the results. The multi-level approach helps the system to deliver results with minimal resource utilization.

Keywords