Research On Analysis Modeling Of Fire Accidents And Its Applications Using Data Mining Technology

Fire disaster has already been recognized as one of serious threats to the survival and development of human being, which shows its own characteristics such as long span of time and space, high frequency of occurrence and heavy losses?With the development of data collection and storage technology, relevant departments would be easier to collect and store huge raw data related with fire.At the same time, theories and methods of spatial data mining(SDM) also provide a new way for fire research.So as a new research view and idea, SDM technology would face unprecedented opportunities for analyzing data of fire?But so far fire research, in SDM research perspectives, is lack of necessary, systematic attention in the domestic and foreign academic research and practical exploration. Based on this background, the thesis aims to introduce SDM technology to fire research, combining theories and methods of SDM with the theory for fire accident analysis. Some innovation achievement has been accomplished, listed below:(1)In view of the present problems among researches for fire accident, the thesis breaks up the limit of these traditional research object, such as a separate building and creatively points out a new ideal for fire accident in SDM research perspective.(2) Theories and methods of location quotient are introduced to calculate the location quotients of fire dange, fire prevention and mitigation capacity, fire vulnerability and fire risk in studied areas? Accoring to corresponding location quotients, ranks of fire dange, fire prevention and mitigation capacity, fire vulnerability and fire risk are evaluated. At the same time, the causes of the fire location quotient differences are analysed.(3) Gravity theory fire accident is put forward innovatively, and it is introduced to analyse the gravity centre traces of fire frequency and direct economic losses at provincial scale in2005-2010. Using SVM forecasting model, gravity center coordinates in the future are predicted with higher precision. (4) Four types of fire concentration index are defined innovatively, respectively named concentration index of fire frequency based on population(FCI1) and GDP (FCI2),and concentration index of fire direct economic losses based on population(FCI3) and GDP (FCI4). Four types of index are calculated out and analysed in2000-2010.(5) Theories and methods of Space Lorentz Curve and Gini coefficient are introduced to research of fire accident. Lorentz Curves of fire frequency vs population are drawn from2000to2010at the provincial level, at the same time, Gini coefficient of each year is calculated out by the method of curve fitting. Theil index and its one-stage decomposition method are used to study the difference of fire accident distribution. This paper takes the six administrative regions and three homogenous regions as investigated regional unit, and measures difference within, total difference and the difference between regions from2000to2010in China with Theil index.(6) Spatial autocorrelation theory have been thoroughly researched and it applied to spatial autocorrelation of fire accident distribution?The methods, including global spatial autocorrelation, local spatial autocorrelation and scatter plot are used to analyse spatial distribution of fire accident, verifying spatial cluster in spatial distribution, determining the specific location of hotspots and its movement trend?(7) The methods of Nearest Neighbor (NN) and Standard Deviation Ellipse (SDE),are used to describe the statistical indices of discrete tendency and central tendency for fire accident cases. Some useful knowledge of spatial distribution is discovered.(8) A new combination method called gray correlation analysis and multiple linear regression method is applied to predict national fire rates. At the same time, theory and method of gray is applied to comprehensive assessment of fire situation in portrait and landscape orientation.(9) Apriori algorithm is systematically discussed, and it is applied to SDM of fire accident cases to discover the rules and knowledge for fire protection.Accoring to association rules related, some hot-spot areas of fire accident can be labeled. In view of the present problems among researches for fire accident, theories and methods of SDM is introduced to this field innovatively, and research framework in SDM research perspective also designed. This paper will provide a new idea and perspective of fire accident research and of value both in academic research and practical application.This paper contains59Figures,85tables,162references.