Throughout this book, we have learned about ensemble learning and explored its applications in many scenarios. In the introductory chapter, we looked at different examples, datasets, and models, and found that there is no single model or technique that performs better than the others. This means that our guard should always be up when dealing with this matter, and hence the analyst has to proceed with extreme caution. The approach of selecting the best model from among the various models means that we reject all of the models whose performance is slightly less than that of the others, and hence a lot of resources are wasted in pursuit of the best model.

In Chapter 7, The General Ensemble Technique, we saw that if we have multiple classifiers with each classifier being better than a random guess, majority voting of the classifiers gives improved performance. We also saw that with a fairly good number of classifiers, the overall accuracy of the majority voting is higher...