Witaningsih , - (2025) METODE JARINGAN SYARAF TIRUAN BACKPROPAGATION UNTUK PREDIKSI KADAR UNSUR HARA DALAM TANAH. Masters thesis, UNIVERSITAS LAMPUNG .
|
Text
ABSTRAK - Witaningsih.pdf Download (191kB) | Preview |
|
![[img]](http://digilib.unila.ac.id/style/images/fileicons/text.png) |
Text
TESIS FULL - Witaningsih.pdf Restricted to Repository staff only Download (5MB) | Request a copy |
|
|
Text
TESIS FULL TANPA BAB PEMBAHASAN -Witaningsih.pdf Download (5MB) | Preview |
Abstract
Ketersediaan nitrogen (N), fosfor (P), dan kalium (K) sangat memengaruhi produktivitas tanaman. Analisis laboratorium untuk mengukur kadar hara bersifat mahal dan memakan waktu, sehingga dibutuhkan metode prediksi yang efisien. Parameter fisik tanah seperti pH, suhu, kelengasan, dan resistansi memiliki keterkaitan dengan kadar hara dan berpotensi menjadi input prediktif. Penelitian ini bertujuan menganalisis hubungan parameter fisik terhadap kadar N, P, dan K, serta membangun sistem forecasting menggunakan Jaringan Syaraf Tiruan (JST). Data diperoleh dari tanah podsolik merah kuning dengan perlakuan pupuk berbeda. Data dinormalisasi, lalu diproses menggunakan JST berarsitektur 4 input, 2 hidden layer (masing-masing 5 node), dan 1 output dilatih menggunakan metode backpropagation dengan berbagai fungsi aktivasi untuk memperoleh konfigurasi optimal. Hasil pelatihan menunjukkan bahwa seluruh parameter fisik memiliki kecenderungan hubungan terhadap kadar hara, meskipun dengan pola yang berbeda. Pelatihan model menunjukkan bahwa nitrogen menghasilkan R2 sebesar 0,9642, fosfor sebesar 1, dan kalium sebesar 0,9996. Pada tahap validasi, diperoleh R2 sebesar 0,7218 untuk nitrogen, 0,6479 untuk fosfor, dan 0,6137 untuk kalium. Model JST yang dikembangkan memiliki akurasi tinggi pada tahap pelatihan serta performa cukup baik saat validasi, sehingga mendukung potensi penerapan sistem prediksi kadar hara berbasis data fisik tanah secara efisien dan tepat guna. Kata kunci: nitrogen, fosfor, kalium, parameter fisik tanah, jaringan syaraf tiruan, prediksi The availability of nitrogen (N), phosphorus (P), and potassium (K) significantly influences crop productivity. Laboratory analysis to determine soil nutrient content is costly and time-consuming, necessitating an efficient predictive approach. Physical soil parameters such as pH, temperature, moisture, and resistance are known to correlate with nutrient levels and hold potential as predictive inputs. This study aims to analyze the relationship between physical parameters and N, P, and K levels, as well as to develop a forecasting system using an Artificial Neural Network (ANN). The data were obtained from yellow- red podzolic soil subjected to different fertilizer treatments. The data were normalized and processed using an ANN architecture consisting of 4 input nodes, 2 hidden layers (each with 5 nodes), and 1 output node, trained using the backpropagation method with various activation functions to achieve optimal configuration. The training results indicate that all physical parameters show a tendency to relate to nutrient levels, although with different patterns. The training phase produced R2 values of 0.9642 for nitrogen, 1 for phosphorus, and 0.9996 for potassium. During validation, the model achieved R2 values of 0.7218 for nitrogen, 0.6479 for phosphorus, and 0.6137 for potassium. The developed ANN model demonstrated high accuracy during training and reasonably good performance during validation, supporting the potential application of soil nutrient prediction systems based on easily measured physical parameters. Keywords: nitrogen, phosphorus, potassium, soil physical parameters, artificial neural network, prediction
| Item Type: | Thesis (Masters) |
|---|---|
| Subjects: | Universitas Lampung > Fakultas Teknik > Prodi Teknik Informatika |
| Divisions: | ?? magister_elektro ?? |
| Depositing User: | UPT . Siswanti |
| Date Deposited: | 05 Nov 2025 02:17 |
| Last Modified: | 05 Nov 2025 02:17 |
| URI: | http://digilib.unila.ac.id/id/eprint/92696 |
Actions (login required)
 |
View Item |