Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2020
- Goals
- Record
- 7975f165-9f60-4224-a05f-0892b838f375
- Published
- 2026-10-30
Universitas Dian Nuswantoro (UDINUS), with researchers Kusuma, Edi Jaya, Shidik, Guruh Fajar, and Pramunendar, Ricardus Anggi, conducted research on optimizing a neural network for breast cancer classification. The activity addressed classification as a data mining technique considered part of supervised learning. It focused on a Backpropagation Neural Network (BPNN), which has been used in several fields to increase human productivity and can provide better, more natural results than other statistical techniques.
The research described an issue in the BPNN learning process: it could produce inefficient synapse weights in each hidden layer. These ineffective weights could affect network performance. To address this issue, the research proposed BPNN optimization using the Nelder Mead method to identify the appearance of breast cancer. The datasets used were the Breast Cancer Coimbra Dataset (BCCD) and the Wisconsin Breast Cancer Dataset (WBCD).
Testing used accuracy and k-fold validation. The testing results presented better performance than the original BPNN. The best average performance for BCCD was reported in the fifth fold, with 76.5217% accuracy. For WBCD, the highest average result was reported in the fourth fold, with 91.1765% average accuracy. The activity therefore frames UDINUS participation in research applying data mining, BPNN optimization, and the Nelder Mead method to breast cancer classification. The source does not provide further information about the research date, institutional unit, funding, implementation setting, or impacts beyond the reported testing results.