A Study on Different Text Representation Methods for the Negative Selection Algorithm.

Unstructured data, such as text, usually have to be structured before standard machine learning classifiers are applied. In such cases, different representation schemes can be used, such as Bag of Words, the Linguistic Inquiry and Word Count (LIWC), Part-of-Speech Tagging (POS Tagging), and others. The Negative Selection Algorithm (NSA) was designed with inspiration in the immune system to solve binary classification problems, more specifically anomaly detection. This paper investigates the performance of various text representation schemes as input to the NSA. Three different datasets and text representation methods are used, and the results are presented in terms of Accuracy and False Positive Rate.