Image of Fine-tuning support vector machine for water change prediction in biofloc

Text

Fine-tuning support vector machine for water change prediction in biofloc

In this study, we address the critical challenge of managing water change frequency in biofloc aquaculture systems, a key factor in maintaining ecological balance and organism health. Our focus is on optimizing Support Vector Machine (SVM) models to predict water change requirements, with a particular emphasis on overcoming class imbalances in datasets and ensuring robust hyperparameter tuning. The aim is to develop a highly accurate SVM model that can determine the optimal timing and quantity for water changes, thus enhancing the efficiency of biofloc systems. This research stands out for its application of SVM models, specifically adjusted to address the unique challenges in biofloc water management through advanced techniques such as Random Over Sampling, Synthetic Minority Over Sampling Technique (SMOTE), Random Under Sampling, and Near Miss for class imbalance correction, and Grid Search, Random Search, and Small Grid Search for hyperparameter optimization. `The evaluation methodology was anchored in Leave-One-Out Cross-Validation (LOOCV), providing a quantitative assessment of model efficacy, training efficiency, and prediction rapidity. The SVM model, fortified by the selected techniques, achieved an impressive 90% accuracy rate across diverse datasets, as substantiated by LOOCV. A noteworthy finding was the efficacy of combining Near Miss with Small Grid Search in data preprocessing and hyperparameter tuning, respectively. This combination was particularly effective in the Near Miss dataset, where it led to a 77.8% reduction in training time compared to the conventional Grid Search approach. The LOOCV revealed that the optimized SVM model achieved an impressive accuracy of 90%, significantly outperforming other predictive methods.

Ketersediaan
TI24/005TI 24/005Prodi Teknik Informatika (Ruang Skripsi dan Tesis)Tersedia
Informasi Detail
Judul Seri
-
No. Panggil
TI 24/005
Penerbit
Jakarta : Universitas Trilogi.,
Deskripsi Fisik
-
Bahasa
English
ISBN/ISSN
-
Klasifikasi
TI
Tipe Isi
text
Tipe Media
-
Tipe Pembawa
-
Edisi
-
Subjek
Information technology
Trilogi University--Information technology
Info Detail Spesifik
-
Pernyataan Tanggungjawab
Versi lain/terkait

Tidak tersedia versi lain

Lampiran Berkas
Tidak Ada Data
Komentar
Anda harus masuk sebelum memberikan komentar