Enhancing Breast Cancer Diagnosis: A Hybrid Approach with Bidirectional LSTM and Variable Size Firefly Algorithm Optimization

Abstract

Breast cancer stands as a significant global health challenge, ranking as the second leading cause of mortality among women. The increasing complexity of timely and accurate remote diagnosis has spurred the need for advanced technological solutions. Breast cancer prediction involves utilizing risk assessment models to identify individuals at higher risk, enabling early detection and personalized treatment strategies. This research meticulously assesses the effectiveness of various long short-term memory (LSTM) classifiers, including simple LSTM, Vanilla LSTM, Stacked LSTM, and Bidirectional LSTM, utilizing a comprehensive breast cancer dataset. Among these, the Bidirectional LSTM emerges as the preferred choice based on a thorough evaluation of accuracy, precision, recall, and F1-Score metrics. In a strategic move to further enhance precision, the Bidirectional LSTM integrates with the variable step-size firefly algorithm (VSSFF). Renowned for dynamically adjusting its step size, VSSFF offers adaptive exploration and exploitation capabilities in optimization tasks. The resulting hybrid model, HVSSFFLSTM, showcases superior performance in breast cancer prediction, suggesting potential applicability across diverse health conditions. Comparative analyses with other models highlight the exceptional accuracy rates of HVSSFFLSTM, achieving 99.78% (training) and 97.37% (testing), precision rates of 99.56% (training) and 97.22% (testing), recall rates of 100% (training) and 98.59% (testing), F1 scores of 99.82% (training) and 97.9% (testing) and specificity of 99.81% (training) and 99.15% (testing). This study not only underscores the adaptability of VSSFF as a valuable optimization tool but also emphasizes the promising prospects of the proposed hybrid model in advancing automated disease analysis. The results indicate its potential beyond breast cancer, suggesting broader applications in various medical domains.