Game Level Automatic Generating System Based On Data Mining And Data Fusion

In the game development, if level design as the focus was designed one by one, a lot of time and capital will be spent,and players stick low degree. At the same time, important data which has important significant in the game design were contained in the game play data.Data mining and data fusion are two analytic processing data,extracting useful knowledge technology, they has certain complementarity on the function.Therefore, A game level automatic generation method based on data mining and data fusion technology is presented in the paper. In the method, game play data were pretreated by boolean logic and rough set theory combination discretion method, and an attribute reduction algorithm based on the information gain was putted for eliminate redundant attributes. Then a decision tree is constructed by ID3algorithm based on attribute reduction, and then use the D-S data fusion algorithm dealing with the data get a reflect the player behavior data. Then the data are inserted into training data set and using decision tree the data is processed to get the difficulty level of the game for players. Finally, the game level parameters modified according to the difficulty level, and according to game level data automatically generate game levels. This paper to push Sokoban game for example doing the experiment, the system is effective to reduce development cost and improve playability.In the paper, the data mining and data fusion technology were researched and experiment,research results are as follows:(1) According to the traditional uniform discretion is the need of human set some parameters processing data, and using the algorithm of decision tree recognition rate is not high, a discretion algorithm based on Boolean logic and the rough set theory was adopted in the paper. The algorithm does not need some human factors and the recognition rate is relatively high, and it can be effective to original data discretion.(2) For the acquisition of game play in the data redundant attributes, the paper need to use after discretion algorithm processing data.Based on the research of existing attribute reduction algorithm and the important of information gain in the ID3decision tree algorithm, an attribute reduction algorithm based on information gain was putted in the paper,and it was effective to slove the data redundancy.(3) According to not use pruning decision tree is too complex and not higher recognition rate, pre-pruning and after-pruning decision tree optimization algorithms were studied in the paper.Through the analysis of the decision tree pre-pruning and after-pruning processing decision tree recognition rate, the algorithm balance the decision tree complexity and accuracy.Based on the full consideration of the complexity of the decision tree and classification accuracy, structure out of a tree as far as possible simple decision tree.(4) The data mining and data fusion technology integration was studied int the paper.According to two kinds of technology in the function complementary characteristics, through the combination of two kinds of technology, improved the model of the acquisition process, but also improve the accuracy of fusion.Through the research of two kinds of technical process of comparison, integration model was setted up, and it can effectively applied to information processing.