Sistem Validasi Dokumen Untuk Transaksi Penjualan Dan Pembelian Unit Alat Berat Menggunakan Convolutional Neural Network

Ahmad, Rahman (2023) Sistem Validasi Dokumen Untuk Transaksi Penjualan Dan Pembelian Unit Alat Berat Menggunakan Convolutional Neural Network. Diploma thesis, Institut Teknologi Sepuluh Nopember.

Text 10311910000040-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (4MB) | Request a copy

Abstract

Proses transaksi penjualan dan pembelian unit alat berat di PT. United Tractors (UT) melibatkan berbagai dokumen atau berkas yang perlu disiapkan. Saat ini proses validasi dokumen hanya dilakukan berdasarkan nama file dokumen yang diunggah ke System Application and Processing (SAP) dimana sering terjadi ketidaksesuaian terhadap syarat keabsahan dokumen. Divisi Differentiation and Digitalization (DAD) UT sedang menjalankan proyek pengembangan sistem SAP untuk proses validasi dokumen secara otomatis sehingga tidak memerlukan operator yang melakukan pengecekan dokumen secara manual. Deep learning merupakan salah satu cabang dari machine learning yang memungkinkan komputer dapat belajar dari data pelatihan. Salah satu algoritma dalam deep learning adalah Convolutional Neural Network (CNN) yang digunakan untuk deteksi objek. Metode CNN yang digunakan pada penelitian dapat berfungsi mengidentifikasi objek-objek yang menjadi syarat keabsahan dokumen. Objek-objek tersebut yaitu tandatangan, materai, stempel pelanggan, dan stempel perusahaan. Hasil dari pendeteksian objek dapat membantu proses validasi dokumen yang sedang dikembangkan oleh Tim Robot Process Automation (RPA) perusahaan. Penelitian ini menggunakan proses crop untuk meningkatkan confidence level dalam mendeteksi objek materai, tanda tangan, stempel pelanggan, dan stempel perusahaan. Hasil pengujian menunjukan model training dengan epoch sebesar 200 memiliki rata-rata persentase akurasi class deteksi objek terbesar yaitu, tanda tangan sebesar 94,14%, materai sebesar 95,89%, stempel perusahaan sebesar 90%, dan stempel pelanggan sebesar 61,72%. Berdasarkan hasil pengujian, penggunaan augmentasi untuk menambah variasi data masukan tidak memberikan peningkatan akurasi yang signifikan dalam deteksi objek terutama objek stempel pelanggan.
=================================================================================================================================
The sales and purchase transaction process for heavy equipment units at PT. United Tractors (UT) involves various documents that need to be prepared. Currently, the document validation process is only carried out based on the file names of the documents uploaded to the System Application and Processing (SAP), which often leads to inconsistencies in meeting the document's validity requirements. The Differentiation and Digitalization (DAD) division at UT is currently running a project to develop the SAP system for automatic document validation, eliminating the need for operators to manually check the documents. Deep learning is one of the branches of machine learning that allows computers to learn from training data. One of the algorithms in deep learning is Convolutional Neural Network (CNN), which is used for object detection. The CNN method used in this research can identify the objects that are the requirements for document validity, such as signatures, stamps, and customer and company seals. The results of object detection can assist the document validation process being developed by the Robot Process Automation (RPA) Team of the company. This research utilizes the crop process to increase the confidence level in detecting objects, namely stamps, signatures, customer seals, and company seals. The testing results show that the training model with 200 epochs has the highest average percentage accuracy in object detection classes, with signatures at 94.14%, stamps at 95.89%, company seals at 90%, and customer seals at 61.72%. Based on the testing results, the use of augmentation to increase input data variations did not provide a significant improvement in accuracy in detecting objects, especially customer seals.

Item Type:	Thesis (Diploma)
Uncontrolled Keywords:	Convolutional Neural Network (CNN), Object Detection, Document,Validation System, Deteksi Objek, Sistem Validasi Dokumen.
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	RAHMAN AHMAD
Date Deposited:	11 Aug 2023 01:36
Last Modified:	11 Aug 2023 01:36
URI:	http://repository.its.ac.id/id/eprint/104226

Actions (login required)

View Item