Behavior choice and experience gaining model on turn-based games with the help of decision theory

Nowadays, the vertiginous progress in information technologies also shows itself in the field of computer games as in all fields. The fun need which is one basic need of humans is provided by means of these games significantly and the growth in sector has made quick steps. It is aimed in this thesis that making a consistent profit account belonging to choice which can be selected in turn-based games in appropriate structure. At the same time, it is aimed to design an artificial intelligence which can learn with the help of this account. The approach we have used here is to create a system which behaves as much as like a person. The main purpose of this is to increase the fun factor and time of the games that can be played in games. The Artificial Intelligence is created in order to succeed this aim that is mentioned. It is studied on getting experience of artificial intelligence and it has been struggled to acquire the properties of making mistake in varying proportions with this experiment which is encountered in daily life frequently. Type of game that is projected while working is the game which the choices are applied as turn-based and every choice made can be measured approximately or completely. While making the choices of the characters' choices that are not controlled by person, it is tried to provide that he/she decides in accordance with the earnings as a result of experiences. It is not aimed to find an artificial intelligence that makes absolute truths, it is aimed to create a structure which there is always a probability to make a mistake as in real life. It is aimed that this structure prefers more useful choices depending on the experiences. It has been benefited from decision theorem for decision making process. But the criteria which have been improved by the researcher have been used instead of the criteria which have been used in decision making process because it will be more appropriate for the new situation. Also; on an application that our method has been tested, calculation of gain value belonging to different types of choice, learning and choice election after learning process will be analyzed. It was observed that this method was able to calculate gain values belonging to projected choices accurately and these calculated values were able to reach wanted goals after being used in learning process.