Symptom Principal Component Analysis (SPCA) for Dimensionality Reduction in Categorical Data: A Case Study on Breast Cancer

The symptom analysis is a promising technique for the expansion of a set of random variables through linear combination over a finite field F2. From the projective space, it is essential to choose the most useful subspaces for the reduction of the dimensionality of the categorical dataset. This paper presents a new method for obtaining the principal components based on symptom analysis, namely, super-symptoms generated from an iterative algorithm based on a set of rules for reducing the dimensionality in categorical data. The proposed method is called Symptom Principal Component Analysis (SPCA), which implies studying and analyzing linear data along with the projective subspaces to get the best results. The dataset includes 101 patients collected at Cancer Oncology Hospital, Baghdad Medical City. The proposed method has shown three major components that affect breast cancer, and it explains 92% of the total variations.