Ensemble Methods In Machine Learning
Building Stronger Performing Machine Learning Models
“Teamwork makes the dream work”. This popular quote is the strange saying I use to recall why we may prefer an ensemble over an individual predictor. Informally, an ensemble is a name used to describe a combination of multiple predictors being used as an individual predictor. This combined quality results in an extremely powerful individual predictor, hence why these models regularly appear in the top solutions of machine learning competitions.
The predictive performance of an ensemble is usually better than that of any one of the constituent algorithms individually. However, some tweaks ought to be made in order to attain this goal. For instance, an ensemble tends to yield better results when the individual predictors in the ensemble are diverse. Therefore, several methods used to create an ensemble seek to decorrelate the individual predictors as much as possible before joining them to the ensemble.
There are many ways to create an ensemble, but three of them are more common than the others. The three methods are:
Bootstrap Aggregation (Bagging)
Bootstrap Aggregation is a machine learning meta-algorithm that involves each model within the ensemble vote with equal weight…
