Forest fire prediction using fuzzy prototypical knowledge discovery

Abstract

An application of Zadeh’s prototype theory in the Knowledge Acquisition process, is presented here, and as a practical example, to define a method for predicting the evolution of the forest fire occurrence-danger rate in INCEND-IA: A KBS for prediction and decision support in fighting against forest fires. This method then allows us to interpret any real cyclical situation using a previously discovered paradigm and define the current period. The FPKD (Fuzzy Prototypical Knowledge Discovery) is presented as a mechanism with the aim of generating Prototypes of Data (A new set of data sufficiently representative to be able to summarize or assimilate the behavior of any of the remaining data); but the concept of prototype is a fuzzy concept and Zadeh’s Theory provides an appropriate framework for its application. Data Mining techniques have been used (decision trees, time series, clustering...). Thus, it is possible to calculate the grade of compatibility of a real situation with the prototypes and define the current period using these affinity values, with the objective of predicting the evolution of the following days