An advanced binary slime mould algorithm for feature subset selection in structural health monitoring data

Feature selection (FS) is an important task for data analysis, pattern classification systems, and data mining applications. In this paper, an advanced version of binary slime mould algorithm (ABSMA) is introduced for feature subset selection to enhance the capability of the original slime mould algorithm (SMA) for processing of measured data collected from monitoring sensors installed on structures. In the first step, structural response signals under ambient vibration are pre-processed according to statistical characteristics for feature extraction. In the second step, extracted features of a structure are reduced using an optimization algorithm to find a minimal subset of salient features by removing noisy, irrelevant and redundant data. Finally, the optimized feature vectors are used as inputs to the surrogate models based on radial basis function neural network (RBFNN). A benchmark dataset of a wooden bridge model is considered as a test example. The results indicate that the proposed ABSMA shows better performance and convergence rate in comparison with four well-known metaheuristic optimizations. Furthermore, it can be concluded that the proposed feature subset selection method has the capability of more than 80% data reduction.