이 논문에서는 공간적 통계기법에 근거한 예측적 공간 데이터 마이닝 방법을 제안하고, 산불위험지역을 예측하는데 적용하였다. 제안된 방법은 조건부 확률과 우도비를 이용한 방법으로 과거 산불발생지역에 대해 산불과 관련된 공간데이터 집합들 사이의 정량적 관계에 의존적인 예측 모델이다. 두 가지 방법을 이용하여 산불위험지역 예측도를 만들고, 각 모델의 예측력을 평가하기 위해 산불위험율(FHR : Forest Fire HazardRate)과 예측률곡선(PRC : Prediction Rate Curve)을 이용하였다. 제안된 두 가지 예측모델의 예측력 비교분석 결과, 우도비 방법이 조건부 확률 방법보다 더 우수한 것으로 나타났다. 이 논문에서 제안된 산불위험지역 예측모델을 이용하여 작성된 산불위험지역 예측도는 산불예방과 산불감시장비 및 인력의 효율적인, 배치 등 산불관리의 효율성을 높이는데 많은 도움을 줄 것으로 기대된다.
이 논문에서는 공간적 통계기법에 근거한 예측적 공간 데이터 마이닝 방법을 제안하고, 산불위험지역을 예측하는데 적용하였다. 제안된 방법은 조건부 확률과 우도비를 이용한 방법으로 과거 산불발생지역에 대해 산불과 관련된 공간데이터 집합들 사이의 정량적 관계에 의존적인 예측 모델이다. 두 가지 방법을 이용하여 산불위험지역 예측도를 만들고, 각 모델의 예측력을 평가하기 위해 산불위험율(FHR : Forest Fire Hazard Rate)과 예측률곡선(PRC : Prediction Rate Curve)을 이용하였다. 제안된 두 가지 예측모델의 예측력 비교분석 결과, 우도비 방법이 조건부 확률 방법보다 더 우수한 것으로 나타났다. 이 논문에서 제안된 산불위험지역 예측모델을 이용하여 작성된 산불위험지역 예측도는 산불예방과 산불감시장비 및 인력의 효율적인, 배치 등 산불관리의 효율성을 높이는데 많은 도움을 줄 것으로 기대된다.
In this paper, we propose two predictive spatial data mining based on spatial statistics and apply for predicting the forest fire hazardous area. These are conditional probability and likelihood ratio methods. In these approaches, the prediction models and estimation procedures are depending un the ...
In this paper, we propose two predictive spatial data mining based on spatial statistics and apply for predicting the forest fire hazardous area. These are conditional probability and likelihood ratio methods. In these approaches, the prediction models and estimation procedures are depending un the basic quantitative relationships of spatial data sets relevant forest fire with respect to selected the past forest fire ignition areas. To make forest fire hazardous area prediction map using the two proposed methods and evaluate the performance of prediction power, we applied a FHR (Forest Fire Hazard Rate) and a PRC (Prediction Rate Curve) respectively. In comparison of the prediction power of the two proposed prediction model, the likelihood ratio method is mort powerful than conditional probability method. The proposed model for prediction of forest fire hazardous area would be helpful to increase the efficiency of forest fire management such as prevention of forest fire occurrence and effective placement of forest fire monitoring equipment and manpower.
In this paper, we propose two predictive spatial data mining based on spatial statistics and apply for predicting the forest fire hazardous area. These are conditional probability and likelihood ratio methods. In these approaches, the prediction models and estimation procedures are depending un the basic quantitative relationships of spatial data sets relevant forest fire with respect to selected the past forest fire ignition areas. To make forest fire hazardous area prediction map using the two proposed methods and evaluate the performance of prediction power, we applied a FHR (Forest Fire Hazard Rate) and a PRC (Prediction Rate Curve) respectively. In comparison of the prediction power of the two proposed prediction model, the likelihood ratio method is mort powerful than conditional probability method. The proposed model for prediction of forest fire hazardous area would be helpful to increase the efficiency of forest fire management such as prevention of forest fire occurrence and effective placement of forest fire monitoring equipment and manpower.
* AI 자동 식별 결과로 적합하지 않은 문장이 있을 수 있으니, 이용에 유의하시기 바랍니다.
문제 정의
Efficient tools for extracting information from geo-spatial data generate and manage large geo-spatial data sets. The focus of this work can be of importance to the organization which own large geo-spatial data sets. Data mining products can be a useful tool in decision-making and planning just as they are cur- rently in the business world.
제안 방법
forest fire hazardous area. And then, predictive power of each model has been evaluated, after carrying out cross validation between the models. Next, forest fire hazardous areas have been mapped using the most effective model.
The geographic database sets for fire ignition prediction analysis were described in section 4. In first, an analysis of the rela tionship of each individual independent attribute with the response attribute was performed to get an idea of the rela tive importance of each attribute in explaining fire ignition. The results of the analysis of relevance of each independent attribute for the data sets corresponding to fire ignition is that forest type, elevation, slope, road networks, farm and building boundaries thematic maps have significant attri butes in the data set corresponding to fire ignition.
The organization which make decisi ons based on large spatial data sets spreads across many domains including ecology and environmental management, public safety, transportation, public health and business. In this study, we focused on the application domain of forest fire prevention where forestry managers are interested in finding spatio-temporal distribution of forest fires and pre dicting forest fire hazardous areas from large spatial/non- spatial data sets such as forest maps, topography maps and fire history data. Forest fire provides a good example to stu dy spatio-temporal representations for GIS applications be cause of its spatio-temporal variability.
spatial data sets. In this study, we proposed and applied two prediction models for spatial data mining based on spa tial statistics. These are conditional probability and like^ lihood ratio methods.
We show that, by analyzing historical data on fire ignition point locations, we can gain the necessary predictive capa bility, making it possible to quantify ignition probability in space. The analysis is performed using inductive approaches in a raster geographic information system (GIS), and it ex plores the information contained in the spatial attributes of the phenomenon.
The forest fire hazardous area prediction model described in this paper provided an effective method for estimation of the degree of forest fire hazard. It is based upon the forestry, topography, human activity attributes, which contribute to fire hazard and risk.
that year. The occurrences play s pivotal role in construc ting prediction models by establishing probabilistic r이ation- ships between the pre-1996 forest fires and the remainder of the input data set. The predictions based on those rela tionships were then evaluated by comparing the estimated hazard classes with the distribution of the forest fire ignition locations that had occurred after 1996, i.
The occurrences play s pivotal role in construc ting prediction models by establishing probabilistic r이ation- ships between the pre-1996 forest fires and the remainder of the input data set. The predictions based on those rela tionships were then evaluated by comparing the estimated hazard classes with the distribution of the forest fire ignition locations that had occurred after 1996, i.e., during the period 1997 to 2001.
The proposed model for prediction of forest fire hazardous area would be helpful to increase the efficiency of forest fire management such as prevention of forest fire occurrence and effective placement of forest fire monitoring equipment and manpower. The ability to quantify ignition risk can be the key to a more informed allocation of fire prevention resour ces.
To make forest fire hazardous area map using the two proposed methods and evaluate the performance of predic tion power, we applied a FHR (Forest Fire Hazard Rate) and a PRC (Prediction Rate Curve) respectively. The FHR for each prediction models was calculated by formula described in section 4.
In these approaches, the prediction mo dels and estimation procedures are depending on the basic quantitative relationships of spatial data sets relevant forest fire with respect to selected the past forest fire ignition areas. To m아ce forest fire hazardous area map using the two pro posed methods and evaluate the performance of prediction power, we applied a FHR (Forest Fire Hazard Rate) and a PRC (Prediction Rate Curve) respectively.
대상 데이터
A key element in the forest fire hazardous area prediction modeling is a forest fire hazard model, which estimates the fire hazard potential based upon forest attributes, forest utilization, and topography. In this study, a forest fire risk prediction model using pre dictive spatial data mining is developed to a forest in the Youngdong region of Kangwon province, Republic of Korea. We show that, by analyzing historical data on fire ignition point locations, we can gain the necessary predictive capa bility, making it possible to quantify ignition probability in space.
The study area to predict forest fire hazardous area is Sam- chok city. The largest forest fire in modem history of Korea occurred in April, 2000 in Samchok city, eastern Korea (Fi gure 3).
성능/효과
We have found out the multiple thematic maps emerged as affective to forest fire occurrence are el evation, slope, forest type, road network, farm and habitat zone condition. The effectiveness of the models estimated and tested and showed acceptable degree of goodness. The proposed model developed would be helpful to increase the efficiency of forest fire management such as detection of for est fire occurrence and effective disposition of forest fire prevention resources.
In first, an analysis of the rela tionship of each individual independent attribute with the response attribute was performed to get an idea of the rela tive importance of each attribute in explaining fire ignition. The results of the analysis of relevance of each independent attribute for the data sets corresponding to fire ignition is that forest type, elevation, slope, road networks, farm and building boundaries thematic maps have significant attri butes in the data set corresponding to fire ignition. However, tree diameter, density, ages and aspect maps have the less significant attributes in the data set.
후속연구
The effectiveness of the models estimated and tested and showed acceptable degree of goodness. The proposed model developed would be helpful to increase the efficiency of forest fire management such as detection of for est fire occurrence and effective disposition of forest fire prevention resources.
참고문헌 (23)
R. H. Guting, An Introduction to Spatial Database Systems, In Very Large Data Bases Jorunal(Publisher : Springer Verlag), October, 1994
S. Shekhar and S. Chawla, Spatial Databases : Issues, Implementation and Trends, 2001
G. Greenman, Turning a map into a cake layer of information, New York Times, Feb., 2000
K. Koperski, J. Adhikary, and J. Han, Spatial Data Mining : Progress and Challenges. In In Workshop on Research Issues on Data Mining and Knowledge Discovery (DMKD'96), 1996
D. Mark, Geographical Information Science : Critical Issues in an Emerging Cross-Disciplinary Research Domain, NSF Workshop, Feb
J. Neter and L. Wasseman, Applied Linear Statistical Models, 4th ed, Irwin
S. M. Weiss and C. A. Kulikowski, Computer Systems that Learn : Classification and Prediction Methods from Statistics, Neural Nets, Machine Learning, and Expert Systems. Morgan Kaufman, 1991
P. Kontkanen, P. Myllym.aki and H. Tirri, Predictive data mining with finite mixtures. Proc. 2nd Int. Conf. on Knowledge Discovery and Data Mining, (KDD'96)
D. M. Skapura, Building Neural Networks, ACM Press, 1996
J. R. Quinlan, Induction of decision trees, Machine Learning
J. R. Quinlan, C4.5 : Programs for Machine Learning, Morgan Kaufmann, 1993
L. Breiman, J. Friedman, R. Olshen, and C. Stone, Classification and Regression Trees, Wadsworth International Group, 1984
M. Mehta, R. Agrawal, and J. Rissanen, SLIQ : A fast scalable classifier for data mining, In Proc. 1996 Int. Conference on Extending Database Technology (EDBT'96), Avignon, France, March, 1996
J. Shafer, R. Agrawal, and M. Mehta, SPRINT : A scalable parallel classifier for data mining, In Proc. 1996 Int. Conf. Very Large Data Bases
M. Kamber, L. Winstone, W. Gong, S. Cheng, and J. Han, Generalization and decision tree induction : Efficient classification in data mining, In Proc. of 1997. Int. Workshop on Research Isues on Data Engineering (RIDE'97)
Jong Gyu Han, Yeon Kwang Yeon, Kwang Hoon Chi and Keun Ho Ryu, Prediction of Forest Fire Hazardous Area Using Predictive Spatial Data Mining, Proc. of Int. Conf. on Information and Knowledge Engineering, pp.348-358, 2002
Lusted, L. B., Introduction to Medical Decision Making : Charles Thomas, Springfield, p.271
Aspinall, P. J. and Hill, A. R., Clinical inferences and decisions-I. Diagnosis and Bayes' theorem : Opthalmic and Physiological Optics, Vol.3, pp.295-304, 1983
Spiegelhalter, D. J.;Knill Jones, R. P., Statistical and Knowledge-based approaches to clinical decision-support systems, with an application in gastroenterology : Journal of the royal Statistical Society, A, Part 1, pp.35-77, 1984
Aspinall, R. J., An inductive modeling procedure based on Bayes' theorem for analysis of pattern in spatial data : Internationl Journal of Geographical Information System, Vol.6(2), pp.105-121, 1992
Reboh, R. andReiter, J., A knowledge-based system for regional mineral resource assessment : Final report, SRI project 4119, p.267, 1983
McCammon, R. B., ProspectorⅡ The redesign of Prospector : AI system in Government, pp.27-31, March, 1989, Washington D.C., pp.88-92, 1989
Reddy, R. K., Bonham-Carter, G. F. and Galley, A. G., Developing a geographic expert system for regional mapping of Volcanogenic Massive Sulphide(VMS)deposit potential : Nonrenewable Resources, Vol.1(2), pp.112-124, 1992
※ AI-Helper는 부적절한 답변을 할 수 있습니다.