Variable Length Black Hole for Optimization and Feature Selection

In the high dimensional space, problem of feature selection (FS) can be regarded as combinatorial optimization with complexity due to huge number candidate features. this article, a novel type meta-heuristic searching based on variable length solution space is proposed in order solve dimensionality issue FS and obtain more optimal results. The algorithm uses original black hole baseline for development. Blackhole assumes fixed which decreases efficiency when features high. Furthermore, subject stagnation single exemplar or selection. Hence, modifies various aspects, namely, it enables decomposing into subset dimensions within each dimension separately selecting an represents corresponding dimension. addition, changing solutions criterion. proposes new concept energy indicates decrease effectiveness time exponential way use replace not effective anymore. designated VLBHO compared particle swarm optimization. approach has increased accuracy from 50% 67% forest cover dataset 38% 80% wine dataset.