The Cascade Improved Model Based Deep Forest for Small-scale Datasets Classification

Abstract

It is very important to classify some small-scale datasets accurately in biology. With the rapid advancement of classification models, support vector machine(SVM), Random Forest(RF), Deep Forest, Convolutional Neural Networks(CNNs) are widely used. However, for small-scale datasets, CNNs always need massive datasets to train. Other methods usually can’t achieve better effects. Therefore, in this paper, a new forest model is proposed to solve the problems in small-scale datasets. It improves the classification performance through integrated learning method. The improved model is different from the primitive model in two important aspects. Firstly, considering the fitting quality of every forest, the standard deviation of some most major features in every forest make up a new feature to be transport in the next cascade layer. Secondly, the sub-layer structure is adapted to the cascade layer to increase the training opportunities. Experiments on five datasets demonstrate that our method has better classification effect than other classification models in the small-scale datasets.