Research On Knowledge Discovery From Uncertain Information Based On Rough Sets And Its Application On Urban Traffic Management

With the rapid development of social economy, continuous expansion of city size, continuous increasing of urban population and vehicles, traffic congestion phenomenon are more and more serious. Traffic congestion is becoming a common social phenomenon. Especially in large cities, traffic congestion has become a key of urban to feature play and sustainable development. As modern urban traffic management is a comprehensive integrated high-tech, any small error could hide significant traffic safety hazard and may cause traffic accidents, result in significant losses. Therefore, information is the future direction of urban traffic management, which apply the latest science and technology, especially information technology to achieve real-time operating state of traffic flow awareness, real-time forecasts and real-time monitoring, and data mining and knowledge discovery is the most important contents. How to manage and process vast amounts of traffic management datas effectively to achieve information management and control on urban traffic is currently a major problem of manager, scientific researcher and engineering staff.Rough sets as a new mathematical tool to deal with uncertain knowledge, compared with others knowledge discovery technology, Rough sets have unique advantages, it need not provide any prior information and knowledge which required processing data collection, through analysis of knowledge reduction and dependency, hidden information or knowledge are discovered and potential rules are revealed. Therefore, rough sets theory can help decision-makers deal with complex systems. This dissertation studied knowledge discovery and decision process framework from uncertain information based on rough sets. Under the basic framework, application theories of knowledge discovery from uncertain information and decision support fields based on rough sets are studied, including discretization, attribute reduction, knowledge discovery, reasoning and interpretation, knowledge discovery from uncertain information and decision-making on urban traffic management based on rough sets theory.The main results of this dissertation are as follows:Through introduction of swarm intelligence optimization theory and rough sets theroy, an improved particle swarm optimization algorithm of continuous attribute descretization is presented. A typical5-group UCI data sets such as breast, iris, wine, glass and heart are adopted to improved particle swarm optimization algorithms. compared with discretization algorithms such as attribute significance based, information entropy and genetic algorithm, the experiment results show that improved discretization algorithm can make minimize loss of information and access to the simplest decision rules from decision-making systems.Considering the minimum attribute reduction NP problem, the concept of discernibility matrix is promoted in this dissertation, and relative discernibility matrix is constructed from perspective of information theory, as conditional entropy to heuristic information, a novel heuristic attribute reduction algorithm is presented, the theoretical analysis and example results show that the algorithm can reduct decision-making information system effectively; Considering large data sets for decision information system, this dissertation introduce concepts of divergent matrix, based on this concept, an improved attribute reduction algorithm which calculate a minimum relative reduction directly based on attribute frequency is constructed, the reduction algorithm can fuse more heuristic information, it will convet repeated divergent logic operation into matrix operation, theoretical analysis and example results demonstrate that this algorithm is superior to other general attribute reduction algorithm, and get minimum relative reduction sets better.Two decision-making rules acquisition algorithms are presented. For complete information system, a global optimizied algorithm for rough sets decision-making rules is presented, the algorithm begin to continuous attributes discretization, based on core attributes, the most important attribute is increased to get the best attribute reduction sets, and finally, all determined and possible decision-making rule sets based on core attribute value reduciton algorithm are acquisizied, calculate methods on evaluation index of certainty and support are given; For incomplete information system, this dissertation introduce concepts of condition attribute matrix and decision attribute matrix, an matrix decision-making rule algorithm based on compatibility relationship from incomplete informaiton system are presented, the algorithm can extract all rules from decision-making system directly without calculating core attributes, so improve algorithm efficiency, which provide a new way for large-scale data sets. Theoretical analysis and example results show that two algorithms are effective.To urban traffic management decision-making issues, this dissertation as rough sets theory a tool, pattern recognition on urban traffic state issues is deeply studied. This dissertation analysized uncertainty factors on urban traffic management, studied the general process and methods of pattern recognition on urban traffic flow state, and calssification knowledge discovery models and algorithms on traffic flow state pattern recognition are peresented, an integrated classification system on traffic flow state pattern from integrated strategy perspective is constructed, and traffic flow state pattern recognition algorithm is presented. Example results show that rough sets theory applied to urban traffic flow state classification knowledge discovery and decision-making have significant application value.Research achievements provide a theoretical basis and practical guidance to urban traffic information management and decision-making.