Game prop personalized recommendation method

Abstract

The present invention relates to a game prop personalized recommendation method, comprising: carrying out extraction, cleaning and processing on data; mapping the data into a low-dimensional space from an input space; classifying the data to partition players into different categories; and recommending props which are most possibly bought by each player for the player. According to the game prop personalized recommendation method disclosed by the present invention, a calculated amount of a recommendation algorithm is greatly reduced, prop recommendation accuracy is significantly improved, and a user experience effect of a network game is promoted.

Description

technical field [0001] The invention relates to the field of online games, in particular to a method for personalized recommendation of game props. Background technique [0002] In existing online games, there are numerous game props. The following three technologies are usually used to realize the personalized recommendation of game props: (1) data dimensionality reduction technology; (2) classification algorithm; (3) recommendation algorithm. Each algorithm technology plays a crucial role in the game item recommendation algorithm. The three algorithms are combined into an algorithm system, and none of them is dispensable. [0003] Among them, the data dimensionality reduction technology needs to solve the complex problem of game data. Because the game data itself has some non-traditional characteristics, such as: game data sources are complex, data types are diverse, data dimensions are high (hundreds or thousands), and data is massive. Therefore, before using the recom...

AI Technical Summary

Problems solved by technology

However, since the props of online games are all virtual items, the data characteristics of online games are very different from those of e-commerce, so moving the recommendation algorithm of e-commerce to online game data will inevitably cause discomfort

The currently public ItemCF algorithm is based on the items (Item) that the user has already purchased. First, calculate the distance matrix between items and then recommend new items (Item) to each user. This calculation method is directly applied to There are the following disadvantages in online game data: (1) There are many types of props in large-scale online games, and the distance matrix between the constructed items is bound to be huge, which will greatly increase the amount of calculation in the future

(2) There are many props in the game that have identical functions and can be substituted for each other. Players who buy such props should be extremely similar, and the props recommended to them should also be similar. The distance between these props with the same function is very large, so the props recommended to players who buy such props are very different

Images