An Improved Algorithm Design and Implementation Based on Unsupervised Learning in Computer Games

At present, the computer games with the intelligent algorithm optimization are always a popular research direction of the artificial intelligence. In this research field, the key concerned technologies and the important resolve are how to continue to gain the final winning rate constantly by independent identification realization, movement optimization and real-time update games based on the unsupervised learning algorithm in the computer chess games. According to the typical artificial intelligence application scene of computer games, an improved algorithm based on the unsupervised learning is designed and proposed in the paper. This algorithm uses alfa-beta search tree to select the best matching pattern and applies the concept of feature learning to make the image recognition innovatively. With the characteristics of the Checkers and the common search methods, it proposes a feature and pattern learning method based on the K-means clustering algorithm. By the experimental data and results, the improved algorithm proposed in the paper can gain and keep a high winning rate in the real games.