The Research on Feature Reduction of Multi-source Heterogeneous Sensor Network

Feature reduction of the multi-source heterogeneous sensor networks is discussed in this paper. First, this paper uses variety distance analytical methods which based on the classification of null value to calculate the interval distribution of existent null value in incomplete information system; Second, a tolerance relation based on interpolated existent null value is proposed due to the study of a binary relation which contains both null value type and non-existent value; Third, the concept of knowledge granularity is introduced, the knowledge granularity measurement method is used to quantify the ability of distinguish between knowledge, and then feature attributes importance are defined from the perspective of granularity calculation. At last, an efficient heuristic reduction algorithm is proposed which based on the attribute importance. This technology takes into account a variety of types of null value at the same time which has an important theoretical and practical significance.