Friction Stir Welding is a feasible choice for joining of Aluminium Metal Matrix Composites (AMMC) over the fusion welding due to the formation of narrow Heat Affected Zone and minimizing the formation of Intermetallic Compounds at weld interface. In this study, a decision tree algorithm-based confusion matrix is developed to classify the AMMC weldments. Initially, the AMMC plates are prepared at five combinations of SiC and B4C reinforcements and welded at three level parameter settings. The weldments are designated as good, partial and bad by analysing the bead height. Also, during welding, the acoustic signals are recorded, and signal analysis is performed using statistical feature extraction. Subsequently, Random Forest algorithm is implemented using the extracted data to yield a confusion matrix to describe the performance of the classifier. Experimental validations are performed by welding the samples at optimal parametric level followed by tensile and microscopic studies. The results emphasize that, the developed classifier resulted in 90.35% classification efficiency by correlating with the experimental validations. Such soft competing algorithms and statistical tools can be used to replicate experimental results by saving energy consumption, resources and material. Copyright © 2019 American Scientific Publishers All rights reserved.