Ensemble-classifier-chain-model

In this paper, we propose a novel computational tool for investigation of anti-inflammatory. First, we encode the anti-inflammatory epitopes in different feature representations, in order to effectively explore more informative feature sets. Then, we use Pseudo (PSE) and Discrete Wavelet Transform (DWT) to convert the variable length coding matrix into the equidimensional features. Also, we select the important feature items based on t-test. Six individual feature sets (DWT-AAC, DWT-PSSM, DWT-PP, PSE-AAC, PSE-PSSM, PSE-PP) can be used to construct various base classifiers. A straightforward way is to integrate different types of features as the input to train a classifier and build a predictive model. Compared with a single learning model, the advantage of ensemble learning model is that it can combine multiple single learning models to form a unified integrated predictor to obtain more accurate, stable and robust results. However, if the feature vector dimension used for integration is too high, which can lead to dimensional disasters; on the other hand, simple integration of different types of features can also result in information redundancy and thereby influence the predictive performance. Finally, we build an ensemble classifier chain model for identifying anti-inflammatory epitopes.