RANCANG BANGUN SISTEM DETEKSI DINI KESULITAN BELAJAR SISWA BERBASIS EXPLAINABLE MACHINE LEARNING DENGAN METODE WATERFALL

Ekta, Pramudya (2026) RANCANG BANGUN SISTEM DETEKSI DINI KESULITAN BELAJAR SISWA BERBASIS EXPLAINABLE MACHINE LEARNING DENGAN METODE WATERFALL. Diploma thesis, UIN Raden Intan Lampung.

	PDF Download (4MB)
	PDF Restricted to Repository staff only Download (4MB)

Abstract

ABSTRAK kesulitan Penelitian ini bertujuan merancang dan membangun sistem deteksi dini belajar siswa berbasis Explainable Machine Learning menggunakan metode Waterfall, dengan algoritma Random Forest untuk klasifikasi risiko dan SHAP sebagai modul penjelas. Sistem dikembangkan melalui pendekatan Research and Development (R&D) dengan arsitektur decoupled Django backend dan React frontend, serta diuji melalui validasi ahli, Black Box Testing, dan UAT di MAN 1 Lampung Timur. Hasil penelitian menunjukkan sistem memperoleh predikat Sangat Baik pada validasi ahli, waktu respons di bawah 5 detik, seluruh fitur berfungsi tanpa bug, akurasi model 92,4% dengan recall 90,8%, serta skor UAT 4,8 dari 5,0 (96%). Sistem terbukti mampu mengotomatiskan prediksi risiko, mengekstrak faktor penyebab secara transparan melalui SHAP, dan memberikan rekomendasi intervensi konkret, sehingga layak diimplementasikan sebagai solusi deteksi dini kesulitan belajar di sekolah. Kata kunci: Early Detection, Learning Difficulties, Explainable Machine Learning, Random Forest, SHAP, Waterfall. ABSTRACT This study aims to design and build an early detection system for student learning difficulties based on Explainable Machine Learning using the Waterfall method, employing the Random Forest algorithm for risk classification and SHAP as an explanatory module. The system was developed through a Research and Development (R&D) approach with a decoupled Django backend and React frontend architecture, and tested through expert validation, Black Box Testing, and UAT at MAN 1 Lampung Timur. The results showed that the system achieved a Very Good rating in expert validation, response time under 5 seconds, all features functioned without bugs, model accuracy of 92,4% with recall of 90,8%, and UAT score of 4.8 out of 5.0 (96%). The system proved capable of automating risk prediction, extracting causal factors transparently through SHAP visualization, and providing concrete intervention recommendations, making it feasible to be implemented as an early detection solution for learning difficulties in schools. Keywords: Early Detection, Learning Difficulties, Explainable Machine Learning, Random Forest, SHAP, Waterfall.

Actions (login required)

View Item