An efficient ensemble framework for outlier detection using bio-inspired algorithm
P. Sharon Femi, Ganesh Vaidyanathan · International Journal of Bio-Inspired Computation · 2022
Outliers are unexpected observations present in data that has to be identified systematically to prevent from catastrophic effects. The detection of outliers plays a crucial role in many application domains, as they provide exciting and critical information about abnormal events. So, in this research work, the bio-inspired ensemble-based outlier detection (BIEOD) is employed to determine the outliers. This ensemble framework combines the sequential and parallel ensembles, thereby creating a hybrid model that utilises the advantages of both feature bagging and XGBoost. The chicken swarm optimisation (CSO) is used to increase the deviation between the outliers and inliers according to the chicken competition in optimising the outlier scores. Finally, the XGBoost predicts the outliers based on the optimised feature set. The obtained results on the benchmark datasets reveal that this framework offers significant advancement over the base learners and other existing ensemble methods of outlier detection.