DOMAIN DENSITY DESCRIPTION BASED INCREMENTAL PATTERN CLASSIFICATION METHOD

摘要

An incremental pattern classification method based on domain density description is provided to determine a specific domain distance function, using support vector learning, to describe a domain of the data, and to obtain the domain density in the domain of the data for describing a discrimination function with small numbers of data. An incremental pattern classification method based on domain density description comprises the steps of: determining a specific domain distance function of showing the center of a domain and distance of data, using support vector learning based on each kind, and describing a domain of the data(S10); obtaining the domain density in the domain of the data, using a maximum likelihood estimating method(S20); multiplying/comparing each domain density description function by/with preliminary probability, and determining the kind of the data having the largest value(S30); and judging whether a newly generated data is obtained in a former step when the new data is added and re-learned(S40).