A Novel Hybrid Dynamic Harris Hawks Optimized Gated Recurrent Unit Approach for Breast Cancer Prediction

Abstract

Abstract The breast cancer (BC) prediction is improved through the machine learning (ML) techniques. In this study, we develop an innovative forecasting framework called the Dynamic Harris Hawks Optimized Gated Recurrent Unit (DHH-GRU) for the prediction of BC. It combines the Gated Recurrent Unit (GRU) and Harris Hawks Optimization (HHO) methods. We gathered data and a training set that included the Wisconsin diagnostic BC (WDBC) dataset, which contains 569 patients with malignant and beginning cases. The collected data were pre-processed using min–max normalization, and important features were extracted by Fast Fourier transform (FFT) and the process of reducing the dimensionality with principal component analysis (PCA). Decimal scaling is employed to equalize the various feature effects. The proposed DHH-GRU technique incorporated the GRU for capturing sequential connections on temporal medical information, and the optimization process, DHH optimization, is utilized. The proposed method's effectiveness is compared and estimated with various existing techniques in terms of log-loss (0.06%), accuracy (98.05%), precision (98.09%), F1-score (98.28%), and recall (98.15%). The proposed DHH-GRU method has a more predictive ability with the sequential dependency in capturing GRU and DHH optimization’s combined behaviour of hunting. This method significantly improved the accuracy of BC prediction.

Keywords

• Machine learning (ML)
• Dynamic Harris Hawks Optimized Gated Recurrent Unit (DHH-GRU)
• Breast cancer (BC)