Aumentando o poder preditivo de classificadores lineares através de particionamento por classe
Abstract
This work describes a new classification technique called P2C - Partitioning to Classify. The main goal is to achieve reasonable classification performances using linear prediction methods, even on datasets with non-linear separable data. The proposed technique, inspired by the division-and-conquer strategy, applies a clustering method on each partition made of samples of the same class. Subsequently, the union among the clusters inside each partition is performed, creating a single partition, where each group can contain linearly separable samples. Then, one or more linear classifiers are trained, according to the number of groups. Experiments performed using datasets with different structural and complexity level indicate the overall performance of the prediction is similar or superior to well-known non-linear classification methods. The main advantages of P2C technique are (i) the need for less effort and computational resources, and (ii) the possibility of treating large datasets due to the ease of parallelization of the steps.