SPIN: A Novel Hybrid Dimensionality Reduction Technique for Cervical Cancer Risk Classification

The number and scale of medical databases is increasingly growing, and sophisticated data mining models may be able to assist physicians and professionals in making more effective and applicable decisions. Cervical cancer is a major type of gynaecological cancer and is amongst the live major malignant cancers in women around the world. Cervical cancer signs are usually undetectable in the early stages. The risk factors are developed due to a number of causes, including the human papillomavirus, sexually transmitted diseases (STDs), and smoking. Dimensionality reduction aids in the removal of redundant or irrelevant features from high-dimensional datasets.This work brings forward a novel hybrid Dimensionality Reduction DR) technique to transform data from higher dimensions to lower-feature subspace. This method combines four major techniques of dimensionality reduction i. e. truncated Singular Value Decomposition (tSVD), Principal Component Analysis(PCA), Independent Component Analysis (ICA), and Non-negative Matrix Factorisation (NMF) and combines the components obtained from each technique into a newer reduced data. The title SPIN hence stands for the significant initials of the base techniques used as mentioned respectively. The proposed method is implemented on the Cervical Cancer Risk dataset. To evaluate performance, the classification for suspected Biopsy examination is done using Decision Tree and random Forest Classifiers, which report an accuracy of 95.283% and 99.057% respectively, which is significantly high as compared to the 98% to 98.67% range present in the recent literatue; even with reduced number of components in lower feature sub-space.