OPTIMIZED RECONSTRUCTION OF BOOLEAN-VALUED INFORMATION SYSTEM FROM AGGREGATES THROUGH GENETIC ALGORITHM BASED ON PROBABILISTIC INITIAL POPULATION SELECTION(O-BISGA)

Abstract

Incomplete Boolean-valued information systems (BIS) provide significant challenges in reconstructing information due to missing or ambiguous components. Whereas the approach of recalculating Boolean-valued Information Systems using the Genetic Algorithm (BISGA) presents a solution by employing aggregate sets and random beginning populations, its reliance on stochastic initialisation typically leads to inefficient convergence and variable accuracy. This study proposes a deterministic, probability-based initial population selection technique to improve BISGA performance. The suggested technique (O-BISGA) distributes 1s to high-probability cells based on weighted averages from row, column, and diagonal aggregates, guaranteeing that chromosomes comply closely to aggregate prerequisites. Experimental findings on BIS matrices (4×4 to 7×7) show a 35% reduction in convergence iterations (8.05 vs. 12.4) compared to BISGA, coupled with increased fitness ratings. The technique reduces randomness-induced inefficiencies while maintaining BISGA's essential limitations, providing a scalable solution for reconstructing incomplete BIS.