Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2025
- Goals
- Record
- dd68eb04-4877-4f11-8849-c4b494284d46
- Published
- 2026-10-30
Universitas Dian Nuswantoro (UDINUS), with the listed authors as activity participants, conducted a study on breast cancer detection using a hybrid classification pipeline. The activity addressed the need for early and accurate diagnosis described in the source and used the Wisconsin Diagnostic Breast Cancer dataset.
The method integrated Hybrid Statistical Feature Selection with unsupervised LSTM-guided feature extraction. Initially, 20 features were selected using Mutual Information, Chi-square, and Pearson Correlation. The training set was balanced using the Synthetic Minority Oversampling Technique to address class imbalance. An LSTM encoder then extracted nonlinear latent features from the selected features. The statistical and latent features were combined through a fusion strategy, followed by re-selection of the top 30 features. Final classification used a Support Vector Machine with an RBF kernel. Evaluation used 5-fold cross-validation and a held-out test set.
The experimental results reported an average training accuracy of 98.13%, an F1-score of 98.13%, and an AUC-ROC of 99.55%. On the held-out test set, the model achieved an accuracy of 99.30%, a precision of 100%, an F1-score of 99.05%, and an AUC-ROC of 0.9973. The source states that the proposed pipeline demonstrated improved generalization and interpretability compared with LightGBM-PSO, DHH-GRU, and ensemble deep networks. The findings highlight the effectiveness of combining statistical selection and LSTM-based latent feature encoding in a balanced classification framework.