Research On Key Technologies And Its Applications Of Neighborhood Granulation And Rough Computing

The rough set theory, which is treated as the mathematical tool for uncertainty analysis, plays an important role in many fields, such as artificial intelligence, data mining, pattern recognition, etc. However, the classical rough set model has the intrinsic limitation which is only suitable for dealing with nominal data. In recent years,researchers have proposed different extensions of rough set theory where neighborhood rough set is one of most important extension models. Neighborhood rough set extends the partition in classical rough set model to the neighborhood covering. This method can handle numerical and nominal data better. With the development of information technology, it presents some fresh challenges to neighborhood rough set theory. For example, how to built the neighborhood rough set model on two universes; how to describe the otherness between neighborhood structure and other granular structures;how to deal with the new problem in the real environment by using the neighborhood rough set theory. We take an extensive study on key technologies of these issues. The contributions of this thesis are as follows.(1) We construct the rough sets model on two universes. However, the classical lower approximation is too strict. The quantitative description by upper approximation is also not available. Hence we propose the variable precision neighborhood rough set model. We define the tolerance neighborhood entropy for dealing with the measurement of incomplete information systems. Finally, the significance of features for decision is also discussed.(2) The otherness exists in different kernel granulation. In order to solve this problem, the chapter four is devoted to the construction of the multi-kernelized granulation rough set model based. Then the properties of multi-kernelized approximation operatiors are discussed in detail. We also apply multi-kernelized approximate quality measure to evaluate the features. On the other hand, there are two essential problems to be addressed in a common rough set model:(1) information granulation;(2) rough approximate. The idea of existing multi-granulation is expressed through the rough approximate. In this study, we propose a novel multi-granulation based on the information granulation. We define open and conservative information granule. Then the definitions of open and conservative muti-granulation entropy are alsogiven respectively. Finally, the effectiveness of muti-granulation entropy-based feature selection method is verified by the experiments.(3) The most traditional feature selection methods focus on the significance of the individual feature which ignores the contribution of the feature in the attribute subset.To deal with this problem, we redefine the redundancy, interdependence and independence of features by using neighborhood entropy. Then the neighborhood entropy-based feature contribution is proposed under the framework of cooperative game. We give the feature a high score if it can improve the ability of the feature subset.The evaluative criteria of features can be formalized as the product of contribution and significance. Finally, the experimental results show that neighborhood entropy-based cooperative game theory model(NECGT) yield better performance than classical ones.(4) Some classical methods are not available for the cold-start problem when the user and the item are new at the same time. In this study, we use neighborhood rough set on two universes(NRSTU) to describe the user and item data structures. The neighborhood lower approximation operator is used for defining the preference rules.Furthermore, the rating table is divided into positive mapping and negative mapping based on the baseline evaluation for the 5 scales rating mechanism. Finally, we give an experimental example to show the details of NRSTU-based preference mining for cold-start problem. The parameters of the model are also discussed. The experimental results show that the proposed method presents an effective solution for preference mining.In conclusion, the neighborhood rough set model is studied from two sides, the neighborhood granulation side and rough approximation side. In the neighborhood granulation aspect, we propose the tolerance neighborhood entropy and the mult-granulation entropy. Then we define the significance of features for decision and the contribution of features by using the concept of the entropy. In the rough approximation aspect, the significance of features for decision is discussed based on multi-kernelized lower approximation. In addition, the neighborhood rough set on two universes is applied to the problem of preference mining.