Combining Models to Improve Classifier Accuracy and Robustness 1
Dean Abbott · 1999
Recent years have shown an explosion in research related to the combination of predictions from individual classification or estimation models, and results have been very promising. By combining predictions, more robust and accurate models are almost guaranteed to be generated without the need for the high-degree of fine-tuning required for singlemodel solutions. Typically, however, the models for the combination process are drawn from the same model family, though this need not be the case. This paper summarizes the current direction of research in combining models, and then demonstrates a process for combining models from diverse algorithm families. Results for two datasets are shown and compared with the most popular methods for combining models within algorithm families.