AcneGrader: An ensemble pruning of the deep learning base models to grade acne

• DOI: 10.1111/srt.13166
• Publication Date: 2022-05-31T17:33:49Z
• Authors: Shuai Liu; Yusi Fan; Meiyu Duan; Yueying Wang; Guoxiong Su; Yanjiao Ren; Lan Huang; Fengfeng Zhou
• Keywords: Pruning; Ensemble Learning; Base (topology); Ensemble forecasting

Abstract Background Acne is one of the most common skin lesions in adolescents. Some severe or inflammatory acne leads to scars, which may have major impacts on patients’ quality life even job prospects. Grading plays an important role diagnosis, and diagnosis made by counting number acne. It a labor‐intensive it easy for dermatologists make mistakes, so very develop automatic methods. Ensemble learning improve prediction results base models, but its time complexity relatively high. The ensemble pruning strategy solve this computational challenge removing redundant models. Materials methods This study proposed novel framework deep models accurately detect grade using images. First, we train multi‐base prune redundancy according performance diversity Then, construct new features training data select previous step. Next, remove further feature selection algorithm. Finally, integrate all classifiers. algorithm was Results experimental showed that pruned achieved accuracy 85.82% dataset, better than existing studies. To verify our method's effectiveness, test method cancer dataset greatly outperform state‐of‐the‐art Conclusion used Our outperforms two datasets, can also reduce complexity.