A technique for Spatial Data Classification Method Using Random Forest based Correlation

Abstract


 
 
 Spatial data mining is a technique mainly used for predictive analytics. Data mining is the method of determining necessary samples from spatial datasets using machine learning methods. But, the existing prediction methods failed to forecast accurately with more accuracy and reduced error. A Random Forest Correlation based Fuzzy classification technique is introduced to improve the spatial data classification and error reduction. The algorithm constructs a random forest and the correlation is measured between the training and testing features. A fuzzy rule based classification is performed for classification into subsets. The proposed method takes forest fire dataset as input and evaluates the parameters such as classification accuracy, space complexity and classification time. The output is a subset of classes of fired and non fired region with enhanced classification accuracy, minimal false-positive rate and reduced time complexity.