This paper proposes a novel ensemble method to improve the performance of binary classification. The proposed method is a non-linear combination of base models and an application of adaptive selection of the most suitable model for each data instance. Ensemble methods, an important type of machine learning technique, have drawn a lot of attention in both academic research and practical applications, and they use multiple single models to construct a hybrid model. A hybrid model generally performs better compared to a single individual model. The proposed approach in this paper based on a hybrid model has been validated on Repeat Buyers Prediction dataset, and the experiment results show up to 18.5% improvement on F1 score, compared to the best individual model. In addition, the proposed method outperforms two other commonly used ensemble methods (Averaging and Stacking) in terms of improved F1 score.

Additional Metadata
Keywords AdaBoost, Adaptive method, Binary classification, Ensemble method, F1 score, Logistic regression, ROC curve, Xgboost
Persistent URL dx.doi.org/10.1109/COMPSAC.2017.49
Conference 41st IEEE Annual Computer Software and Applications Conference, COMPSAC 2017
Citation
Fan, X. (Xing), Lung, C.H, & Ajila, S. (2017). An Adaptive Diversity-Based Ensemble Method for Binary Classification. In Proceedings - International Computer Software and Applications Conference (pp. 822–827). doi:10.1109/COMPSAC.2017.49