Deep neural network method to study computer game

Computer game is the main research direction in the field of artificial intelligence, the research object of this computer gameselection for card games of incomplete information game dou landlord, aiming at the problem of traditional modeling methods need to manually extract the features of, through the analysis of the principle of rival modeling algorithm and characteristics, this paper proposes a neural network and decision tree based on depth assessment modeling method of rivals, The tedious manual feature extraction process is avoided. The experimental results show that the overall winning chip is higher than that of the agents using static evaluation method, which improves the level of the system game to a certain extent.