Factor-relevance Tree And Its Application In Traffic Accidents Analysis

Searching for relationship amang datas is one of the main purposes of data m-ining. Association rules mining is a basic way of data mining, which aims at relation-ship between items and reveals the latent rules hidden in the database. But sometimes these so detailed rules can not meet all the needs of users, who may also requires the relativity between attributes. Usually it's necessary to analyse relativity between variables in a complicated system, which help simplify the framework of the system, discover the leading factors and contradiction that effect the behavior of system, and evaluate the charactor of system behavior that help make a decision. Regarding a system variable as a attribute, the relativity between attributes is converted into the relativity between attributes.In a broad sense, association rules mining and relativity between attributes both be-long to relationship-mining which is a branch of data mining. In this thesis, an item in database is considered as a value of an attribute. Therefore an association rule reflects the connection of values of attributes. So association rules mining, relativity between attributes and relativity between attribute-sets are closely related and promoted one by one structurally. If a complete relationship-mining is required, these three level of study must be done one by one.On purpose of the complete and progressive relationship-mining, this thesis builds A new relationship-mining model named factor-relevance tree based on association rules mining method and the entropy theory. This model is specially used for mining the relations between datas in a large database, which shows the gradation of relativity in structure. As well, the algorithm with the model is in term of top-down pattern.In the end of this thesis, the model and its algorithm is applied to the study on road traffic accidents, successfully find out the leading factors that effect the severity of traffic accidents and master a few valuable rules that hidden in the traffic accidents database. The result could be a supplementary to the relevant department, help them take measures in order to reduce casualties and property loss caused by traffic accdents timely and effectively.