An instance of Satisfiability Problem learnt with Instance based, Decision trees, Naive Bayes

Abstract

Though it is not recommended for mining the data sets produced by any closed-form solution, our approach is very much aiming at the standard satisfiability (SAT) or assignment problem in terms of searching for the existence of learning models and their performance using machine learning algorithms. Here we claim that we do not apply the analytical or theoretical method in the sense it is not based on conventional or well-established minimization rules or circuit synthesis in mind but the application of computational codes for prediction for the given Boolean circuits. The main objectives are to get classification on mapping Boolean literal values to some predefined functional output and to get maximum accuracy of such classifiers. Here we have enumerated the results up to 20 bits majority operator as an instance for constructing the learning models based on machine learning algorithms. The similarity and significance of the results are shown and thus it may open the possibility for extensibility and generalizability to other types of Boolean circuits. The maximum average performance is found to be 80.72% of accuracy and 74.81% of precision based on the selected set of classifiers.