June Scientific Discussion of the Undergraduate Statistics Program: A Closer Look at Quantile Regression Forest / Diskusi Ilmiah Bulan Juni Prodi S1 Statistika: Mengenal Lebih Dekat Quantile Regression Forest

Bengkulu, June 26, 2025 – The Statistics Study Program of the Faculty of Mathematics and Natural Sciences (FMIPA), University of Bengkulu, held another session of its Monthly Scientific Discussion on Thursday, June 26, 2025, at the Undergraduate Statistics Program Room. The event was attended by lecturers from the Statistics Program, with Winalia Agwil, S.Si., M.Si., serving as the presenter, delivering a talk on “Regression Tree, Random Forest, and Quantile Regression Forest.”

In her presentation, Ms. Winalia explained that conventional linear regression methods are often inadequate when dealing with nonlinear data. As an alternative, the Regression Tree was introduced as a nonparametric approach to regression modeling. This method constructs a decision tree by splitting data based on the optimal variables and cut-off points to minimize the Residual Sum of Squares (RSS).

She further elaborated on the Random Forest technique, which employs bagging to combine multiple regression trees, resulting in more stable predictions. The discussion culminated with an in-depth explanation of Quantile Regression Forest (QRF)—an extension of Random Forest capable of providing prediction intervals by retaining the full distribution of response values at each terminal node of the tree.

The discussion session was lively, with participants posing various questions, including the distinction between regression and classification in supervised learning, how the regression tree algorithm determines the best split, how to interpret QRF results, and the potential for overfitting in training data.

The event concluded with a summary emphasizing that nonlinear approaches such as Regression Tree, Random Forest, and Quantile Regression Forest are highly valuable for handling complex data and hold significant potential in research and applications of machine learning-based analysis. It is hoped that this scientific discussion series will continue to enrich the knowledge and insight of the academic community in the Undergraduate Statistics Program in facing the challenges of modern data analysis.

Bengkulu, 26 Juni 2025 – Prodi Statistika FMIPA Universitas Bengkulu kembali menyelenggarakan kegiatan Diskusi Ilmiah Bulanan pada Kamis, 26 Juni 2025 bertempat di Ruang Prodi S1 Statistika. Kegiatan ini dihadiri oleh dosen Prodi statistika dengan penyaji Winalia Agwil, S.Si., M.Si., yang membawakan topik “Regression Tree, Random Forest, and Quantile Regression Forest”.

Dalam paparannya, Ibu Winalia menjelaskan bahwa metode regresi linier konvensional sering kali tidak memadai untuk menangani data yang bersifat nonlinier. Sebagai alternatif, Regression Tree diperkenalkan sebagai pendekatan nonparametrik dalam pemodelan regresi. Metode ini membentuk pohon keputusan dengan membagi data berdasarkan variabel dan titik potong terbaik untuk meminimalkan Residual Sum of Squares (RSS).

Lebih lanjut, dijelaskan pula mengenai Random Forest, yang menggunakan teknik bagging untuk menggabungkan beberapa regression tree dan menghasilkan prediksi yang lebih stabil. Materi diskusi mencapai puncaknya saat pembahasan tentang Quantile Regression Forest (QRF)—sebuah pengembangan dari Random Forest yang mampu memberikan interval prediksi melalui penyimpanan distribusi penuh nilai respon di setiap simpul terminal pohon.

Sesi diskusi berlangsung aktif dengan berbagai pertanyaan dari peserta, di antaranya mengenai perbedaan regression dan classification dalam supervised learning, cara kerja algoritma regression tree dalam menentukan split terbaik, interpretasi hasil QRF, hingga potensi overfitting pada data pelatihan.

Kegiatan ini ditutup dengan kesimpulan bahwa pendekatan nonlinier seperti Regression Tree dan Random Forest, termasuk Quantile Regression Forest, sangat berguna dalam menangani data kompleks dan memiliki potensi besar dalam penelitian dan penerapan analisis berbasis machine learning. Diharapkan kegiatan diskusi ilmiah ini dapat terus memperkaya wawasan sivitas akademika Prodi S1 Statistika dalam menghadapi tantangan analisis data modern.