Desain Model Artificial Neural Network-Particle Swarm Optimization (ANN-PSO) Untuk Prediksi Produktivitas Padi

Ramadhani, Litakuni Windriya (2026) Desain Model Artificial Neural Network-Particle Swarm Optimization (ANN-PSO) Untuk Prediksi Produktivitas Padi. Diploma thesis, Institut Teknologi Sepuluh Nopember.

Text 2042221119_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Produktivitas padi dipengaruhi oleh kondisi tanaman, tanah, cuaca, dan vegetasi sehingga memerlukan pemodelan berbasis data. Penelitian ini bertujuan merancang dan mengevaluasi model Artificial Neural Network-Particle Swarm Optimization (ANN-PSO) untuk memprediksi produktivitas padi. Penelitian menggunakan 514 sampel petak berukuran 1 × 1 m dengan sepuluh fitur masukan, meliputi parameter tanaman, tanah, vegetasi, dan curah hujan. Sampel dikumpulkan secara langsung dari area persawahan di Desa Kedungwaras, Kecamatan Modo, Kabupaten Lamongan, pada satu musim tanam untuk menggambarkan kondisi pengamatan pada lokasi penelitian. Produktivitas aktual dalam kg/m² digunakan sebagai target. Data dibagi menjadi 80% data pelatihan dan 20% data pengujian, kemudian dinormalisasi pada rentang 0,1-0,9. Model terpilih menggunakan arsitektur 10-6-1. PSO dijalankan dengan 50 partikel dan 100 iterasi menggunakan MSE sebagai fungsi fitness. Solusi terbaik PSO digunakan sebagai weight dan bias awal pada proses fine-tuning ANN menggunakan trainbr dengan Bayesian Regularization. Pada data pengujian, model menghasilkan MSE 0,004385, MAE 0,0550 kg/m², dan R² 0,4680. Kinerja model sedikit lebih baik dibandingkan ANN baseline. Model selanjutnya diterapkan dalam GUI berbasis MATLAB App Designer untuk mendukung pelatihan, pengujian, visualisasi, dan prediksi. Hasil penelitian menunjukkan bahwa model ANN-PSO dapat menjadi salah satu pendekatan dalam prediksi produktivitas padi berdasarkan kondisi dan dataset penelitian yang digunakan.
=======================================================================================================================================
Rice productivity is influenced by various factors, including crop conditions, soil, weather, and vegetation, which require data-driven modeling approaches. This study aims to design and evaluate an Artificial Neural Network-Particle Swarm Optimization (ANN-PSO) model for rice productivity prediction. The study used 514 plot samples with a size of 1 × 1 m and ten input features, including crop, soil, vegetation, and rainfall parameters. The samples were collected directly from a rice field area in Kedungwaras Village, Modo District, Lamongan Regency, during a single planting season to represent the observed conditions at the research location. Actual productivity in kg/m² was used as the target output. The data were divided into 80% training data and 20% testing data, then normalized within the range of 0.1-0.9. The selected model used a 10-6-1 architecture. PSO was performed using 50 particles and 100 iterations with MSE as the fitness function. The best PSO solution was used as the initial weight and bias for ANN fine-tuning using trainbr with Bayesian Regularization. On the testing data, the model achieved an MSE of 0.004385, MAE of 0.0550 kg/m², and R² of 0.4680. The model performance was slightly better than the ANN baseline. The developed model was further implemented in a MATLAB App Designer-based GUI to support training, testing, visualization, and prediction processes. The results indicate that the ANN-PSO model can serve as one of the approaches for rice productivity prediction based on the conditions and dataset used in this study.

Item Type:	Thesis (Diploma)
Uncontrolled Keywords:	ANN-PSO, prediksi produktivitas padi, NDVI, machine learning, dan akuisisi data, rice productivity prediction, data acquisition.
Subjects:	T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.84 Heuristic algorithms.
Divisions:	Faculty of Vocational > Instrumentation Engineering
Depositing User:	Litakuni Windriya Ramadhani
Date Deposited:	14 Sep 2026 02:31
Last Modified:	14 Sep 2026 02:31
URI:	http://repository.its.ac.id/id/eprint/144480

Actions (login required)

View Item