Measurement classification using hybrid weighted Naive Bayes

This paper presents an algorithm for classifying measurement variables within airborne measurement data files collected by NASA. The proposed solution utilizes a combination of decision tree and Naive Bayes classifiers. In order to mitigate the independence assumption of Naive Bayes, we apply a weight vector to the feature set based on each feature's role in the classification process. The Analytic Hierarchy Process is selected to calculate the weight vector, after an investigation of various weight calculation techniques. The assessment of the algorithm with recent NASA data shows that the algorithm delivers robust results, and exceeds the performance expectation in the presence of inconsistencies and inaccuracies among measurement data.