Framework for Pre-processing, Feature $\boldsymbol{-}$ extraction, and Specialisation of Nominal Data

There have always been many challenges to understand and use nominal data as effectively as numerical data. The judgment-based scaling of nominal data is not as efficient as a formula-based transformation of numerical data where mean, mode, median, and standard deviation are calculated. In this paper, we propose a novel framework to process nominal data by transforming nominal attributes into equivalent statistically significant values for any given data domain. The transformation process has three stages. The three stages consist of addition of domain knowledge, preservation of collected attributes/features, and evaluation of numerical values by identifying co-relations and similarities of the features. These three stages in the framework will help improve the nominal data processing and will address issues related to knowledge transformation such as adding existing knowledge to the data domain, data loss such as rejecting less relevant attributes, and feature transformation such as transforming the values while processing nominal data.