<![CDATA[Abstrak - Proposal akademik merupakan dokumen penting yang menjadi dasar penilaian kelayakan penelitian mahasiswa sebelum memasuki tahap tugas akhir. Proses evaluasi proposal yang masih dilakukan secara manual sering menghadapi kendala berupa subjektivitas penilaian, waktu pemeriksaan yang lama, serta meningkatnya beban kerja akademik. Penelitian ini bertujuan mengembangkan ScanPro, sebuah sistem pendukung keputusan berbasis web yang mengintegrasikan Lexical Analyzer berbasis Deterministic Finite Automata (DFA) dan algoritma Decision Tree C4.5 untuk mengevaluasi kelayakan proposal akademik secara otomatis. Sistem melakukan ekstraksi teks dari berbagai format dokumen menggunakan PDF.js, Mammoth.js, dan Tesseract.js OCR, kemudian menganalisis karakter secara leksikal untuk menghasilkan tujuh fitur kuantitatif akademik yang digunakan sebagai atribut klasifikasi. Dataset penelitian terdiri dari 150 dokumen proposal yang dikategorikan ke dalam kelas Layak, Layak dengan Revisi, dan Tidak Layak. Model dievaluasi menggunakan 30 data uji independen dengan metrik Accuracy, Precision, dan Recall. Hasil pengujian menunjukkan bahwa sistem mampu mencapai akurasi sebesar 86,67%, macro precision sebesar 86,40%, dan macro recall sebesar 86,40%. Selain menghasilkan klasifikasi kelayakan, sistem juga mampu memberikan rekomendasi perbaikan secara otomatis berdasarkan aturan pohon keputusan yang tidak terpenuhi. Kebaruan penelitian ini terletak pada integrasi DFA sebagai mekanisme ekstraksi fitur leksikal berbasis automata dengan algoritma C4.5 dalam arsitektur client-side yang menjaga privasi dokumen, mengurangi latensi pemrosesan, dan mendukung evaluasi proposal akademik secara objektif, transparan, serta efisien. Kata kunci : Deterministic Finite Automata; C4.5; Klasifikasi Proposal; Lexical Analyzer; Sistem Pendukung Keputusan; Abstract - Academic proposals are fundamental documents that serve as a prerequisite for students before conducting their final research projects or theses. Conventional proposal evaluation processes often face challenges such as lengthy review times, subjective assessments, and increasing workloads for academic reviewers. This study proposes ScanPro, a web-based intelligent decision support system that integrates a Deterministic Finite Automata (DFA)-based Lexical Analyzer with the C4.5 Decision Tree algorithm to automate the feasibility assessment of academic proposals. The system performs cross-format text extraction using PDF.js, Mammoth.js, and Tesseract.js OCR, followed by lexical analysis to generate seven quantitative academic features, including keyword density, academic term ratio, structural completeness, numerical consistency, lexical error rate, lexical diversity, and text length validation. These features are subsequently utilized as classification attributes for the C4.5 model. The experimental dataset consisted of 150 academic proposal documents categorized into three classes: Eligible, Eligible with Revision, and Not Eligible. The model was evaluated using 30 independent testing documents and assessed through Accuracy, Precision, and Recall metrics. Experimental results demonstrated that the proposed system achieved an accuracy of 86.67%, macro precision of 86.40%, and macro recall of 86.40%. Furthermore, the system is capable of generating automated revision recommendations based on unmet decision tree rules, thereby providing actionable feedback to users. The novelty of this research lies in the integration of DFA-based lexical feature extraction and the C4.5 classification algorithm within a fully client-side architecture that preserves document privacy, minimizes processing latency, and supports objective, transparent, and efficient academic proposal evaluation. Keywords: Deterministic Finite Automata; C4.5; Academic Proposal Evaluation; Lexical Analyzer; Decision Support System;]]>
Copyrights © 2026
