Applications of Neurocomputing (KC-LSTM) in Prediction Fire - A Comprehensive Study and Understanding

this paper present predictor model combination between the advantage of Knowledge Constriction (KC) and Long Short-Term Memory (LSTM to find the burned area. The dataset of this case syudy collected from five types of sensors, namely Temperature Sensor (Tmp36), Wind Speed Sensor (010C), Relative Humidity Sensor (RH5X), Rain Sensor (WR2-RFC A553100), and Direct Current (Acudc240). The predictor model consists of four stages, the first stage called pre-process, include (1) check the missing values (2) split dataset into multi-intervals and remove the duplicate; the second stage build knowledge constraints, then split the data set into two parts: training and testing. We'll create the predictor with the training data set while putting the based on five cross validations. The third stage, building predictor dependent on the neurocomputing (LSTM) and knowledge constriction called (4F-DLT), the four-stage using the evaluations measure. As a results the 4F-DLT consider a pragmatic predictor for fire forest area reduce the computation and time at the same time increase the accuracy.