Rancang Bangun Frontend dan Model Prediksi Produktivitas Pada Sistem Taksasi Produksi Kelapa Sawit

Muhammad, Wafi Zaki Hanif (2026) Rancang Bangun Frontend dan Model Prediksi Produktivitas Pada Sistem Taksasi Produksi Kelapa Sawit. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221039-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Estimasi produktivitas kelapa sawit merupakan aspek manajerial yang krusial untuk mengoptimalkan rantai pasok industri perkebunan. Namun, pengembangan sistem peramalan kerap terhambat oleh data starvation serta ketiadaan platform terintegrasi yang mampu memvisualisasikan kondisi kebun. Penelitian ini bertujuan membangun aplikasi berbasis web dengan arsitektur Modular Monolith yang mengintegrasikan modul pemetaan spasial (Tree Counting), visualisasi 3 dimensi (Volume Estimate), dan model machine learning prediktif. Sistem dibangun menggunakan framework Next.js, sedangkan tantangan keterbatasan data dianalisis menggunakan algoritma regresi XGBoost melalui evaluasi komparatif antara data asli (baseline) dan augmentasi data sintetis. Hasil pengujian fungsionalitas menunjukkan aplikasi mampu memvalidasi skenario Black-Box Testing dengan tingkat keberhasilan 100%. Pengujian non-fungsional membuktikan sistem mampu mempertahankan stabilitas frame rate dan waktu muat di bawah 30 detik pada kondisi jaringan throttle 3G. Pada aspek kecerdasan buatan, eksperimen menunjukkan bahwa penambahan data sintetis justru memicu terjadinya severe overfitting dan kontradiksi logika kausalitas fitur pada analisis interpretasi SHAP. Oleh karena itu, skenario XGBoost baseline ditetapkan sebagai model terbaik karena memiliki konsistensi penalaran agronomi yang paling valid, dengan capaian nilai RMSE sebesar 2,361 ± 0,536, MAE sebesar 1,833 ± 0,348, MAPE sebesar 17,01% ± 2,88, serta nilai koefisien determinasi (R^2) mencapai 0,8129 ± 0,1299. Integrasi arsitektur web modern dan pemodelan yang konsisten ini berhasil merealisasikan purwarupa manajemen taksasi kelapa sawit yang andal dan adaptif.
========================================================================================================================
=========
Oil palm productivity estimation is a crucial managerial aspect for optimizing the plantation industry supply chain. However, the development of forecasting systems is often hindered by data starvation and the lack of an integrated platform capable of visualizing plantation conditions. This study aims to build a web-based application using a Modular Monolith architecture that integrates spatial mapping (Tree Counting), 3D visualization (Volume Estimate), and a predictive machine learning model. The system was built using the Next.js framework, while the data limitation challenge was analyzed using the XGBoost regression algorithm through a comparative evaluation between baseline (real data) and synthetic data augmentation. The functional testing results indicate that the application successfully validated all scenarios via Black-Box Testing with a 100% success rate. Non-functional testing proves that the system maintains frame rate stability and page load times under 30 seconds even under 3G network throttling conditions. In the artificial intelligence aspect, experiments revealed that adding synthetic data triggered severe overfitting and logical contradictions in feature causality within the SHAP interpretation analysis. Therefore, the XGBoost baseline scenario was selected as the definitive best model due to its valid agronomic reasoning consistency, achieving an RMSE of 2.361 ± 0.536, MAE of 1.833 ± 0.348, MAPE of 17.01% ± 2.88, and a coefficient of determination (R^2) of 0.8129 ± 0.1299. The integration of a modern web architecture and consistent modeling successfully realized a reliable and adaptive oil palm yield estimation prototype.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Estimasi produktivitas, Machine learning, Modular monolith, Kelapa sawit, XGBoost. Machine learning, Modular monolith, Palm oil, Productivity estimation, XGBoost.
Subjects:	T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Wafi Zaki Hanif
Date Deposited:	27 Jul 2026 01:41
Last Modified:	27 Jul 2026 01:41
URI:	http://repository.its.ac.id/id/eprint/137793

Actions (login required)

View Item