Improving performance of random forests for a particular value of outcome by adding chosen features

Choosing features to improve a performance of a particular algorithm is a difficult question. Currently here is PCA, which is hard to understand (although it can be used out-of-the-box), is not easy to interpret and requires centralizing and scaling of features. In addition, it does not allow to improve prediction performance for a particular outcome (if its accuracy is lower than for others or it has a particular importance). My method enables to use features without preprocessing. Therefore a resulting prediction is easy to explain.