20.500.12556/RUL-175396 Napovedovanje koncentracij fitoplanktonskih pigmentov v Tržaškem zalivu s pomočjo metod strojnega učenja Forecasting phytoplankton pigment concentrations in the Gulf of Trieste with machine learning techniques Fitoplankton predstavlja pomemben pokazatelj ekološkega stanja morskega okolja. Ker so in-situ meritve težko dostopne in časovno zahtevne za analizo, smo raziskali alternativne metode za ocenjevanje koncentracij fitoplanktonskih pigmentov na podlagi različnih okoljskih parametrov ter raziskali možnost njihove napovedi v Tržaškem zalivu z uporabo modelov strojnega učenja. V sodelovanju s strokovnjaki s področja smo izbrali niz ekološko relevantnih značilk za oblikovanje vhodnega nabora podatkov, katerega smo ustrezno predprocesirali. V raziskavi smo uporabili različne regresijske modele ter nevronske mreže, pri čemer smo zaradi omejenega števila in-situ meritev pigmentov uporabili strategijo učenja s prenosom znanja. Model smo najprej trenirali na satelitskih meritvah klorofila a in ga nato dodatno prilagodili na in-situ meritvah. V ta namen smo uporabili večslojne perceptrone (MLP) in konvolucijske nevronske mreže (CNN). Uporaba učenja s prenosom znanja je pokazala izboljšanje napovedne natančnosti, kar kaže na njen potencial za izboljšanje ocenjevanja fitoplanktonskih pigmentov v podatkovno omejenih morskih okoljih. Phytoplankton represents an important indicator of the ecological state of the marine environment. As in-situ measurements are difficult to obtain and time-consuming to analyze, we explored alternative methods to estimate phytoplankton pigment concentrations from other environmental parameters and investigated the feasibility of estimating phytoplankton pigment concentrations in the Gulf of Trieste using machine learning models. A set of ecologically relevant features was selected in collaboration with domain experts to construct the input dataset, which was preprocessed including appropriate handling of missing values. The study employs basic regression models and neural networks, and since real pigment measurements are scarce, we employed a transfer learning strategy: we pre-trained a model on satellite-derived chlorophyll-a data and fine-tuned it with ground measurements. For this purpose, multilayer perceptrons (MLPs) and convolutional neural networks (CNNs) were applied. The application of transfer learning led to an improvement in predictive performance, demonstrating its potential to enhance phytoplankton pigment estimation in data-limited marine environments. strojno učenje okoljske znanosti učenje s prenosom znanja koncentracije fitoplanktonskih pigmentov machine learning environmental sciences transfer learning concentrations of phytoplankton pigments true false false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2025-10-25 08:15:05 2025-10-25 08:15:12 2025-12-10 03:48:07 0000-00-00 00:00:00 2025 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 1970-01-01 004.8:502 155326 254424579 20983.pdf 20983.pdf 1 D0955A792792BCCA8EB1591A4FA22E01 3fd03cf49c1623eb128b6a8db20da2f481f2a109916dcb79486a750451c40b64 ad17ffee-b169-11f0-9328-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=220925 Fakulteta za matematiko in fiziko 0 0 0