Max-voting, which is generally used for classification problems, is one of the simplest ways of combining predictions from multiple machine learning algorithms.
In max-voting, each base model makes a prediction and votes for each sample. Only the sample class with the highest votes is included in the final predictive class.
For example, let's say we have an online survey, in which consumers answer a question in a five-level Likert scale. We can assume that a few consumers will provide a rating of five, while others will provide a rating of four, and so on. If a majority, say more than 50% of the consumers, provide a rating of four, then the final rating is taken as four. In this example, taking the final rating as four is similar to taking a mode for all of the ratings.
