Parallel Compound Index Arrays for Flexible Classification

The goal of this research is to develop a new type of classification method that can be used to infer the value of any attribute of the dataset it has been trained on (given the values of the other attributes) without the need to retrain its model. For this reason, a new classifier named Parallel Compound Index Arrays (PCIA) is introduced that aims to achieve this by creating an association between the known attribute values of the training data and the appropriate training samples, which can be used to quickly find the sample that is the most similar to any given input data. Furthermore, the proposed implementation applies parallel computing procedures in order to enable the classifier process larger volumes of data. The efficacy of the proposed classifier is shown using 10 benchmark datasets, compared to other contemporary classification methods.