Effectiveness of principal component analysis in breast cancer diagnosis: An empirical and visualization analysis

Abstract

With the development of computer science technology, it is increasingly common to apply computer science technology to the medical field, and one of the typical examples is to predict the probability of cancer cells recurrence in cancer patients. However, there is still a lack of unified explanation for the impact of the principal component analysis (PCA) dimensionality reduction method used in prediction on the final experimental results. Therefore, the research topic of this paper is the effect of PCA dimensionality reduction method on the accuracy of cancer prediction. The methodology in this paper is as follows: First, patient data were collected, based on a usable machine learning repositor. It has 569 instances, which contains both malignant and benign tumors. Second, exploratory data analysis is performed on the data, detecting outliers and discovering existing correlated data. Then, the comparative method is used to explore the effectiveness of PCA dimensionality reduction method on the prediction accuracy by observing and analyzing tables and images. It is found that the PCA dimensionality reduction method plays a positive role for boosting the prediction accuracy, but the PCA method does not significantly improve the prediction accuracy for the data that has been processed and the dimension is not high.