Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2021
- Goals
- Record
- a1c0d6ed-945e-4cc7-97ac-21339cd8b913
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted research on attribute selection for Random Forest classification in an unbalanced diabetes dataset. The activity involved researchers Rachmawanto, Eko Hari; Ignatius Moses Setiadi, De Rosal; Rijati, Nova; Susanto, Ajib; Wahyu Mulyono, Ibnu Utomo; and Rahmalan, Hidayah. The research addressed diabetes classification in a context where much previous research had focused on one dataset with limited attributes. It examined the role of attribute selection in the classification process, particularly when using the Random Forest method.
The activity compared two datasets with different types and numbers of attributes. The research considered the bagging feature of Random Forest as reliable for the learning process on an unbalanced dataset. Testing identified age, bs _ fast, bs _pp, plasma_r, plasma_f, hba1c, and type/class as attributes that played an important role in improving the accuracy of diabetes mellitus classification. The research further reported that data cleaning and attribute selection, combined with the Random Forest method, produced 100% accuracy on Abel Vika's Diabetes dataset while using a relatively small number of trees. This result was validated using the k-fold cross-validation method.
Through this activity, UDINUS contributed research focused on diabetes classification, attribute selection, data cleaning, and validation of classification results. The activity is relevant to the stated goal through its focus on health-related research and diabetes mellitus classification. The evidence records a research-based analytical activity conducted by UDINUS and presents its reported findings without extending beyond the described datasets, attributes, method, and validation approach.