Repository of Faculty of Science
Not a member yet
1747 research outputs found
Sort by
Analysis of the Impact of Hyperparameters on the Performance of Machine Learning Models
No full textU ovom radu istražena je primjena modela strojnog učenja, s posebnim naglaskom na neuronske mreže, za detekciju korisnika telekomunikacijskih usluga koji potencijalno namjeravaju promijeniti operatera. Korišten je neuravnotežen skup podataka Telco, pri čemu su primijenjeni algoritmi za uzorkovanje radi smanjenja utjecaja klasne neuravnoteženosti. Optimizacija modela provedena je korištenjem Hyperband algoritma za pretragu hiperparametara te metodom odzivnih površina za finu prilagodbu ključnih parametara. Provedena je i PRIZMA analiza relevantne literature te replikacija odabranog istraživanja radi usporedbe rezultata. Dobiveni rezultati ukazuju na poboljšanje performansi modela u odnosu na prethodne pristupe, pri čemu su analizirani i utjecaji različitih tehnika obrade podataka.This paper explores the application of machine learning models, with a particular focus on neural networks, to detect telecommunications service users who potentially intend to switch operators. An unbalanced Telco dataset was used, with sampling algorithms applied to reduce the impact of class imbalance. Model optimization was performed using the Hyperband algorithm for hyperparameter search and the response surface method for fine-tuning key parameters. A PRIZMA analysis of the relevant literature was also performed, and replication of selected research was performed to compare the results. The results obtained indicate an improvement in model performance compared to previous approaches, while the impacts of different data processing techniques were also analyzed
Comparison of computer analysis of electricity production of two offshore wind turbines
No full textZavršni rad pruža usporedbu računalne analize proizvodnje električne energije dviju pučinskih vjetroturbina. U radu su opisane osnovne vrste vjetroturbina, provedena računalna analiza pučinske vjetroturbine s kružnom stacionarnom domenom te uspoređeni rezultati s vjetroturbinom s pravokutnom stacionarnom domenom.The thesis provides a comparison of the computational analysis of electricity production for two offshore wind turbines. It describes the basic types of wind turbines, presents a computational analysis of an offshore wind turbine with a circular stationary domain, and compares the results with those of a wind turbine with a rectangular stationary domain
Physics of Friction: Adapting Teaching to Different Learning Styles
No full textOvaj rad bavi se prilagodbom nastave fizike različitim stilovima učenja u okviru teme trenja. Detaljno je objašnjena fizika trenja, uključujući osnovne zakone i primjene ovog fenomena u svakodnevnom životu. Stilovi učenja, vizualni, auditivni i kinestetički, igraju ključnu ulogu u načinu na koji učenici usvajaju znanja. Rad analizira pristupe podučavanju trenja uz pomoć osnovnoškolskog udžbenika te nudi preporuke kako se nastavni sadržaji mogu prilagoditi svakom stilu učenja. Kroz primjere interaktivnih metoda i kreativnih rješenja, naglašava se važnost mijenjanja i prilagođavanja nastavnih metoda kako bi se povećala učinkovitost učenja. Integracija različitih stilova učenja unutar jedne lekcije može značajno unaprijediti proces učenja i razumijevanja fizikalnih pojava, a time i potaknuti dublje razumijevanje prirodnih zakona.This work deals with the adaptation of physics teaching to different learning styles on the subject of friction. The physics of friction is explained in detail, including the fundamental laws and real-life applications of this phenomenon. The learning styles, visual, auditory, and kinaesthetic, play a crucial role in how students absorb knowledge. This paper analyzes the approaches to teaching friction from a middle school textbook and provides recommendations on how to tailor teaching materials to each learning style. Examples of interactive methods and creative solutions are used to emphasize the importance of adapting and modifying teaching methods to increase learning effectiveness. Integrating different learning styles into a single lesson can significantly enhance the learning process and deepen students' understanding of physics concepts
The impact of dietary phytoestrogens on human health
No full textBiljke mogu sadržavati biološki aktivne spojeve, a jedni od njih su fitoestrogeni, nesteroidni polifenoli sposobni oponašati ili modulirati djelovanje estrogena. Glavne skupine fitoestrogena su flavonoidni i neflavonoidni, čiji je izvor širok raspon biljnih namirnica koji uključuje žitarice, mahunarke, sjemenke i začine. Fitoestrogeni djeluju vezivanjem za estrogenske receptore, što može dovesti do blagotvornih učinaka poput smanjenja ili prevencije simptoma menopauze i osteoporoze, kao i smanjenje rizika od hormonski ovisnih karcinoma. Također mogu pozitivno utjecati na kardiovaskularni, živčani i imunološki sustav. Međutim, moguće su i negativne posljedice, uključujući djelovanje kao endokrini disruptori koji mogu utjecati na razvoj reproduktivnog sustava, fertilnost i funkciju štitne žlijezde. Uz sve veću popularnost biljne prehrane bogate fitoestrogenima, ovi učinci bi mogli doći do većeg izražaja. Ovaj rad nastoji pružiti jasnu sliku o trenutno dostupnim saznanjima o fitoestrogenima, uključujući njihove izvore, mehanizme i učinke djelovanja, kao i njihovu potencijalnu relevantnost u kontekstu povezanosti prehrane i zdravlja.Plants may contain biologically active compounds, including phytoestrogens, non-steroidal polyphenols that can mimic or modulate the effects of estrogens. The most important groups of phytoestrogens are flavonoids and non-flavonoids, which come from a variety of plant sources, including grains, legumes, seeds and spices. Phytoestrogens exert their effects by binding to estrogen receptors. This can lead to beneficial outcomes such as the reduction or prevention of menopausal symptoms and osteoporosis, as well as a decreased risk of hormone-dependent cancers. They can also have a positive effect on the cardiovascular, nervous and immune systems. However, negative effects are also possible, including their role as endocrine disruptors, which can affect reproductive system development, fertility and thyroid function. With the increasing popularity of plant-based diets rich in phytoestrogens, these effects may become even more pronounced. This paper aims to provide a clear overview of the current state of knowledge on phytoestrogens, including their sources, mechanisms, effects and potential importance in relation to diet and health
SQL related misconceptions among computer science students
No full textSQL programski jezik specifičan je za područje rada s podacima baza podataka. Ispravno korištenje i dobro poznavanje SQL jezika su često u temeljima za donošenje ispravnih poslovnih odluka. Ovaj rad predstavlja analizu istraživanja koje ispituje poznavanje SQL programskog jezika među studentima Informatike Prirodoslovno-matematičkog fakulteta u Splitu te postojanje različitih miskoncepcija vezanih uz kreiranje SQL upita nad relacijskim bazama podataka.SQL programming language is specific to the field of working with database data. Proper use and thorough knowledge of SQL are often fundamental for making sound business decisions. This paper presents an analysis of research examining the knowledge of the SQL programming language among students of Informatics at the Faculty of Science in Split, as well as the existence of various misconceptions related to creating SQL queries on relational databases
Differential Equations in Dynamical Systems
No full textCilj ovog rada je doći do rješenja raznih diferencijalnih jednadžbi u linearnim sustavima. Koristeći se raznim tehnikama i alatima za rješavanje upoznajemo se s rješenjima homogenih i negomogenih linearnih sustava. Fundamentalni teorem egzistencije i jedinstvenosti nam potvrđuje postojanje rješenja problema početne vrijednosti.The goal of this work is to find solutions to various differential equations in linear systems. Using various techniques and tools for solving, we get to know the solutions of homogeneous and nonhomogeneous linear systems. The Fundamental Existence-Uniqueness Theorem confirms the existence of a solution to the initial value problem
Creating a light game on the Arduino platform
No full textOvaj rad prikazuje razvoj funkcionalnog sustava igre križić-kružić koristeći Arduino platformu, senzore i LED tehnologiju. Rad započinje pregledom Arduino platforme, s fokusom na model ploče Mega 2560, njegove komponente te postupak programiranja Arduino pločica. Nastavlja se detaljnom analizom senzora, posebno senzora svjetlosti poput fotootpornika, fotodioda i fototranzistora. Također se razmatra LED tehnologija, uključujući princip rada, karakteristike, vrste LED dioda te njihove prednosti u odnosu na klasične izvore svjetlosti. Igra križić-kružić analizira se kroz njena pravila i strategije koje igrači mogu koristiti. Praktični dio rada opisuje zahtjeve sustava, korištene dijelove, proces izrade te implementaciju igre koristeći Arduino, uz vizualni prikaz sustava i izvorni kod. Rad završava procjenom mogućnosti za poboljšanje i proširenje funkcionalnosti sustava.This paper presents the development of a functional Tic-Tac-Toe game system using the Arduino platform, sensors, and LED technology. The paper begins with an overview of the Arduino platform, focusing on the Mega 2560 model board, its components, and the process of programming Arduino boards. It continues with a detailed exploration of sensors, particularly light sensors such as photoresistors, photodiodes, and phototransistors. LED technology is also examined, covering the operating principles, characteristics, types of LEDs, and their advantages over traditional light sources. The Tic-Tac-Toe game is analyzed through its rules and player strategies. The practical section of the paper outlines the system requirements, components, design process, and implementation of the game using Arduino, with a visual representation of the system and source code. The paper concludes with an assessment of potential improvements and expansions of the system's functionality
Fragile X chromosome
No full textFragilni kromosom X je dominantno svojstvo i jedan od glavnih uzroka mentalne retardacije u muškaraca. Uzrok navedenog poremećaja je nedostatak ili mutacija gena FMR1. Normalni gen FMR1 u svojoj 5’-netranslacijskoj regiji sadrži 6-50 kopija tripleta CGG, a mutirani gen 200-4000 kopija CGG tripleta. Veliki broj metiliranih tripleta CGG utišava transkripciju gena FMR1 i dovodi do nemogućnosti stvaranja proteina Fmr1. Kod žena simptomi najčešće nisu izraženi, a kod muškaraca se uočavaju brojna obilježja fenotipa: izduženo lice s naglašenim ušima, hipotonija mišića, uvećani testisi i mentalna retardacija. Lijek za fragilni kromosom X nije otkriven, ali postoje brojne terapije i tretmani koji mogu ublažiti simptome i olakšati život pojedinaca.Fragile X chromosome is X-linked dominant disorder and is one of the main causes of mental retardation in boys. The cause of this disorder is a deficiency or mutation in the FMR1 gene. The normal FMR1 gene contains 6-50 copies of CGG triplets in its 5’-untraslated region, while the mutated gene contains 200-4000 copies of CGG triplets. A high amount of methylated CGG triplets silences FMR1 gene transcription and leads to the inability to produce the Fmr1 protein. Symptoms are not expressed in girls, while in boys, numerous characteristics of phenotype are observed: elongated face with prominent ears, muscle hypotonia, enlarged testicles and mental retardation. Cure for fragile X syndrome is not discovered, but there are numerous therapies and treatments that can alleviate symptoms and make life easier for individuals
Uklanjanje zamućenja kretanja na slikama ljudskog lica korištenjem dubokog učenja
No full textFacial image degradation by motion blur is a problem that continues to occur with the increased usage and presence of cameras in our daily lives. Whether the purpose of image capturing is personal use or further higher level computer vision tasks, they both require deblurred facial image. This thesis covers the latest literature and describes every aspect of facial deblurring problem. The importance of training dataset is tested by training different models on different facial datasets. After the trained models are tested and compared in detail the discovered conclusions were used for further development and testing of deblurring models with the emphasis on common problems and limitations.Degradacija slike lica zbog zamućenja kretanja problem je koji se i dalje javlja s povećanom upotrebom i prisutnošću kamera u našem svakodnevnom životu. Bilo da je svrha snimanja slike osobna upotreba ili daljnji zadaci više razine računalnog vida, oboje zahtijevaju uklanjanje zamućenja slike lica. Ovaj rad pokriva najnoviju literaturu i opisuje svaki aspekt problema uklanjanja zamućenja lica. Važnost skupa podataka za treniranje testirana je treniranjem različitih modela na različitim skupovima podataka lica. Nakon što su uvježbani modeli testirani i detaljno uspoređeni, otkriveni zaključci korišteni su za daljnji razvoj i testiranje modela uklanjanja zamućenja s naglaskom na uobičajene probleme i ograničenja
Modeling a Relational Database from Non-relational Data Sources
No full textU ovom radu opisana je baza podataka za pohranu informacija o studijima i kolegijima koji se izvode na Prirodoslovno-matematičkom fakultetu u Splitu, kao i JSON u SQL pretvarač koji služi za pohranu podataka iz postojeće NoSQL baze podataka u JSON formatu u novostvorenu relacijsku bazu podataka. Također, rad uključuje skriptu koja uspoređuje vrijeme izvršavanja pet upita na SQL i NoSQL bazi podataka, čime se analiziraju performanse oba modela. Baza podataka izrađena je korištenjem Microsoft SQL Server Management Studio 20, dok je pretvarač razvijen u programskom jeziku Python koristeći uređivač Visual Studio Code.This thesis describes a database designed to store information about the study programs and courses offered at the Faculty of Science, University of Split, as well as a JSON to SQL converter that transfers data from the existing NoSQL database in JSON format to a newly created relational database. Additionally, the thesis includes a script that compares the execution time of five queries on both the SQL and NoSQL databases, providing an analysis of the performance of both models. The database was created using Microsoft SQL Server Management Studio 20, while the converter was developed in Python using the Visual Studio Code editor