1. Long-term temperature prediction with hybrid autoencoder algorithmsJorge Pérez-Aracil, Dušan Fister, C. M. Marina, César Peláez-Rodriguez, L. Cornejo-Bueno, P. A. Gutiérrez, Matteo Giuliani, A. Castelleti, Sancho Salcedo-Sanz, 2024, izvirni znanstveni članek Opis: This paper proposes two hybrid approaches based on Autoencoders (AEs) for long-term temperature prediction. The first algorithm comprises an AE trained to learn temperature patterns, which is then linked to a second AE, used to detect possible anomalies and provide a final temperature prediction. The second proposed approach involves training an AE and then using the resulting latent space as input of a neural network, which will provide the final prediction output. Both approaches are tested in long-term air temperature prediction in European cities: seven European locations where major heat waves occurred have been considered. The longterm temperature prediction for the entire year of the heatwave events has been analysed. Results show that the proposed approaches can obtain accurate long-term (up to 4 weeks) temperature prediction, improving Persistence and Climatology in the benchmark models compared. In heatwave periods, where the persistence of the temperature is extremely high, our approach beat the persistence operator in three locations and works similarly in the rest of the cases, showing the potential of this AE-based method for long-term temperature prediction.
Ključne besede: autoencoder, temperature prediction, hybrid models, heatwave
Objavljeno v DKUM: 29.01.2025; Ogledov: 0; Prenosov: 2
 Celotno besedilo (1,82 MB) |
2. A genetic algorithm based ESC model to handle the unknown initial conditions of state of charge for lithium ion battery cellKristijan Korez, Dušan Fister, Riko Šafarič, 2025, izvirni znanstveni članek Opis: Classic enhanced self-correcting battery equivalent models require proper model parameters and initial conditions such as the initial state of charge for its unbiased functioning. Obtaining parameters is often conducted by optimization using evolutionary algorithms. Obtaining the initial state of charge is often conducted by measurements, which can be burdensome in practice. Incorrect initial conditions can introduce bias, leading to long-term drift and inaccurate state of charge readings. To address this, we propose two simple and efficient equivalent model frameworks that are optimized by a genetic algorithm and are able to determine the initial conditions autonomously. The first framework applies the feedback loop mechanism that gradually with time corrects the externally given initial condition that is originally a biased arbitrary value within a certain domain. The second framework applies the genetic algorithm to search for an unbiased estimate of the initial condition. Long-term experiments have demonstrated that these frameworks do not deviate from controlled benchmarks with known initial conditions. Additionally, our experiments have shown that all implemented models significantly outperformed the well-known ampere-hour coulomb counter integration method, which is prone to drift over time and the extended Kalman filter, that acted with bias.
Ključne besede: enhanced self-correcting model, state of charge estimation, lithium-ion cell parameter identification
Objavljeno v DKUM: 08.01.2025; Ogledov: 0; Prenosov: 4
Celotno besedilo (5,96 MB) |
3. Regulacija lebdenja lahke žogice s fpga : diplomsko deloAlen Jakopič, 2024, diplomsko delo Opis: V diplomski nalogi sem raziskal uporabo FPGA platforme za regulacijo sistema lebdenja lahke žogice. Namen dela je bil ustvariti izobraževalni model, ki pokaže lastnosti PID regulacije in prednosti uporabe FPGA pri regulaciji. Uporabil sem PID regulacijo za doseganje stabilnosti in hitrosti odziva. Rezultati so pokazali, da sistem učinkovito sledi referenčni vrednosti in hitro reagira na spremembe. Sistem je stabilen, brez statičnega pogreška. Kljub nekaterim izzivom z motnjami je regulacija uspešna. Nadaljnje delo bi lahko izboljšalo odzivnost na motnje in natančnost sistema.
Ključne besede: FPGA, PID regulacija, lebdenje, Dewesoft
Objavljeno v DKUM: 14.10.2024; Ogledov: 0; Prenosov: 27
Celotno besedilo (5,34 MB) |
4. Bike sharing and cable car demand forecasting using machine learning and deep learning multivariate time series approachesCésar Peláez-Rodriguez, Jorge Pérez-Aracil, Dušan Fister, Ricardo Torres- López, Sancho Salcedo-Sanz, 2024, izvirni znanstveni članek Ključne besede: cities green mobility, bike sharing demand prediction, cable car demand prediction, machine learning, deep learning
Objavljeno v DKUM: 22.08.2024; Ogledov: 76; Prenosov: 9
Celotno besedilo (4,33 MB) |
5. Toward an economy of wellbeing : the economic impact of the Welsh healthcare sectorTimotej Jagrič, Christine Elisabeth Brown, Dušan Fister, Oliver Darlington, Kathryn Ashton, Mariana Dyakova, Mark Bellis, Vita Jagrič, 2022, izvirni znanstveni članek Opis: Population health and wellbeing is both a result, as well as a driver, of economic development and prosperity on global, European, national and sub-national (local) levels. Wales, one of the four United Kingdom (UK) nations, has shown a long-term commitment to sustainable development and achieving prosperity for all, providing a good example of both national and sub-national level, which can be useful for other European countries and regions. In this paper, the economic importance of the healthcare sector to the Welsh economy is explored. We use a large number of data sources for the UK and Welsh economy to derive an economic model for 2017. We estimate output, income, employment, value-added, and import multipliers of the healthcare sector. Results suggest that the healthcare sector has an above average contribution in four explored economic aspects of the Welsh economy (output, income, employment, value-added), according to its impact on the surrounding economic ecosystem. Also, it is below average regarding leaking through imports. The multipliers' values offer empirical evidence when deciding on alternative policy actions. Such actions can be used as a stimulus for encouraging regional development and post-COVID economic recovery. Our study refers to the Welsh healthcare sector's economic impact as a whole. Therefore, we suggest investigating the economic impact of individual healthcare providers in the future.
Ključne besede: input-output analysis, healthcare sector, Wales, impact analysis, economy of wellbeing
Objavljeno v DKUM: 17.06.2024; Ogledov: 181; Prenosov: 17
Celotno besedilo (544,71 KB)
Gradivo ima več datotek! Več... |
6. Time series numerical association rule mining variants in smart agricultureIztok Fister, Dušan Fister, Iztok Fister, Vili Podgorelec, Sancho Salcedo-Sanz, 2023, izvirni znanstveni članek Ključne besede: association rule mining, smart agriculture, optimization, evolutionary algotihms, internet of things
Objavljeno v DKUM: 12.06.2024; Ogledov: 121; Prenosov: 11
Celotno besedilo (1,49 MB)
Gradivo ima več datotek! Več... |
7. |
8. Uporaba umetne inteligence pri upravljanju portfelja delnicDušan Fister, 2022, doktorska disertacija Opis: Izziv dela predstavlja snovanje, načrtovanje in praktična izvedba avtomatiziranega trgovalnega sistema, ki neodvisno in brez posredovanja uporabnikov sprejema in izvaja trgovalne odločitve. Jedro trgovalnega sistema predstavlja trgovalna strategija, ki spremlja pretekle ter aktualne podatke borznih kotacij, izvaja tehnično analizo in, če je tega sposobna, se prilagaja sprotnim razmeram na finančnih trgih. Obravnavamo dve skupini trgovalnih strategij, klasične, ki niso sposobne sprotnega prilagajanja niti učenja, in dve trgovalni strategiji na osnovi naprednih algoritmov umetne inteligence, eno izmed njih predstavnico umetnih nevronskih mrež najnovejše tretje generacije. Izvedemo obširna simulacijska eksperimentiranja na osnovi nemškega delniškega trga v zadnjih desetih letih, zasnujemo in izvedemo pa tudi eksperimentiranja na namenski strojni opremi, ki močno pohitri kompleksnost časovnega izvajanja, ter eksperimentiranja na analognem elektronskem vezju, s pomočjo katerega se podrobno seznanimo z načinom propagiranja informacij umetnih nevronskih mrež tretje generacije. Rezultati eksperimentov prinašajo tako vsebinske kot tehnične ugotovitve, najpomembnejšo med njimi, da se enoten model ki hkrati trguje z večjim številom finančnih instrumentov obnaša podobno kot kopica posamično prilagojenih modelov na točno določen finančni instrument, kakor tudi novo ugotovljene izkušnje vezane na propagiranje in izrabo najnovejše generacije umetnih nevronskih mrež.
Ključne besede: umetna inteligenca, portfelj delnic, umetne nevronske mreže, mehanski trgovalni sistem
Objavljeno v DKUM: 14.11.2022; Ogledov: 835; Prenosov: 200
Celotno besedilo (5,58 MB) |
9. Algoritmično trgovanje kriptovalut : diplomsko deloJan Podkoritnik, 2022, diplomsko delo Opis: V nalogi smo predstavili trg kriptovalut, splošno teorijo trgovanja in tehnične analize. Opisali smo algoritmično trgovanje in pripadajoče strategije. Implementirali smo aplikacijo, ki omogoča avtomatizirano trgovanje in testiranje izbranih strategij na podlagi preteklih podatkov. Izbrali smo nekaj tehničnih indikatorjev, implementirali strategijo trgovanja in poskušali na podlagi izvedenih testov s pomočjo aplikacije ugotoviti, ali je mogoče biti dobičkonosen. Na koncu smo predstavili analizo rezultatov, pridobljenih s testiranjem.
Ključne besede: kriptovalute, algoritmično trgovanje, C#, tehnična analiza
Objavljeno v DKUM: 25.10.2022; Ogledov: 559; Prenosov: 104
Celotno besedilo (2,96 MB) |
10. Razvoj naprednega adaptivnega regulatorja za mehatronske sistemeDušan Fister, 2017, magistrsko delo Opis: V magistrski nalogi predstavljamo, opisujemo ter razlagamo princip delovanja nelinearnega naprednega adaptivnega hitrostnega regulatorja, ki smo ga izdelali za potrebe napredne regulacije na enoosnem robotu. Tega krmilimo z algoritmom evolucijskih strategij, ki je sposoben dinamičnega iskanja rešitev, kjer se vrednost funkcije uspešnosti spreminja s časom. Pri tem smo vrednost funkcije uspešnosti napovedovali s pomočjo nevronske mreže (angl. Artificial Neural Network, krajše ANN). Predlagano metodo smo testirali na realnem laboratorijskem robotskem sistemu z eno stopnjo prostosti (angl. one degree of freedom, krajše 1 D.O.F.) in ugotovili, da je primeren za regulacijo v realnem času (odzivni čas 1-5 ms). Izvedena je bila primerjava z linearnim PI-hitrostnim regulatorjem, rezultati pa so pokazali uspešnejše delovanje nelinearnega hitrostnega regulatorja.
Ključne besede: online-regulacija, identifikacija, evalvacija, optimizacija, realni čas
Objavljeno v DKUM: 01.09.2017; Ogledov: 1735; Prenosov: 173
Celotno besedilo (2,36 MB) |
