Research On Intelligent Road-network Selection Method Based On Cases Reasoning

Cartographic generalization is influenced greatly by complex thinking of human being, besides,it also exists much uncertainty, so using traditional math model to solve the problem has encounteredmore and more challenges. Cartographic scholars have got a common view that cartographicgeneralization must be intelligent. The past intelligent generalization methods can’t be termed as thereal “intelligence” because of the lack of self-learning property and reasoning ability.The basis of intelligence is knowledge, and the kernel is the reasoning system. Paying attentionto the knowledge while ignoring the reasoning system has lead to the low intelligent level in theprocess of the intelligent cartographic generalization. Therefore, reasoning system is the kernel ofimproving the intelligent level of cartographic generalization.This paper takes road network selection as example, and explores methods which are closer tothe human learning and common cognition. By introducing achievements of reasoning based on casein machine learning, an intelligent selection method of road network based on case learning is putforward, and some special researches have been made on reasoning system. This method takes theroad network selection results of cartographic experts as cases, and computer can obtain theknowledge(rules or case models) from the cases automatically by using learning algorithm(reasoningsystem), which is used to solve the selection of the same or similar road network. This method withefficient reasoning system has the ability of identifying the existing knowledge, obtaining newknowledge, improving the property and self-perfecting, which provides a available way for thedevelopment of intelligent cartographic generalization, and also provides the theory and technologyhelp for cartographic generalization and other intelligence research.This paper makes a major research on the describe of road network case, designment andmanagement of case and case-base, reasoning system of analogy learning, reasoning system ofinductive learning, noise control of case-base.The followings are the main researches of this paper:1. Designing and constructing the model of road network case. A new describe model of roadnetwork information is put forward on the basis of analyzing the restrict factors in the road networkselection, which is aimed at the problem that the information of road network can’t supply enoughinformation supports to the cartographic generalization. Taking the above points as basis, casedescribe and express structure are studied in the process of road network selection. The designmentand management rules of the case of the road network selection are put forward from the view ofcartographic generalization, which form the foundation of the intelligent selection of the roadnetwork based on case learning.2. An intelligent road-network selection method using Cases Based Reasoning (CBR) is put forward in the intelligent selection of road network. The process frame of selection is designed indetail, and some deep researches about the foundation of analogy reasoning system are made. Theprinciples and approaches of the simplification of the case-base and case flood are put forward in theprocess of road network selection, which is combined with the characteristics of road network andbasic principles of selection. The case models with the ability of generalization, which are producedautomatically from the expert case-base, can be used to guide the same or similar selection of roadnetwork.3. An intelligent selection method based on case inductive learning is put forward. Through theresearch of the inductive reasoning system, some inductive and reasoning case rules are made fromcase-base by using decision tree algorithm, which regularize the “invisible experience” of the expert.Then the computer can obtain the expert experience from expert case-base automatically, and theselection of the same or similar road network can be realized.4. The tactics of reducing noise are researched combined with pre-noise reducing and post-noisereducing. As to the negative influence on the reasoning results of the wrong cases in the case-base,firstly, cases are filtered preliminarily by using the cartographic generalization knowledge in order toreject the wrong cases violated the basic principles; then cases in the surplus case-base are comparedand analyzed, and the small probability and the unusual cases should be rejected, which can improvethe quality of the case-base and improve the effect of computer learning.