Uporaba umetne inteligence za izdelavo urnikov dežurstev v zdravstveni ustanovi

Starič, Peter (avtor), Demšar, Jure (mentor)

Zadovoljstvo zaposlenih je zelo pomemben vidik v delovnem okolju. K temu prispevajo tudi obvezna dežurstva, za katera želimo, da so skladna z željami osebja. Ker trenuten sistem izbiranja in dodeljevanja dežurstev v določeni zdravstveni ustanovi ne izpolnjuje pričakovanj, smo se odločili razviti sistem, ki bo to izboljšal. Namesto priprave lastnega algoritma smo pri generiranju urnika preizkusili uporabnost umetne inteligence, kjer smo pozornost namenili produktom podjetja OpenAI, in sicer modelom GPT-3.5-turbo, GPT-4.1 in GPT-5 ter pogovornemu robotu ChatGPT. Modele smo testirali na enakih vhodnih podatkih in jih ovrednotili na podlagi njihovega izhoda, kjer smo preverjali enakomernost razporeditve dela in upoštevanje želja ter dnevov počitka. Pogovorni robot je imel drugačno nalogo, saj je moral pripraviti kodo, ki generira urnik. Ustreznost kode smo ocenili z enakimi vhodnimi podatki kot pri modelih. Ugotovili smo, da se pri generiranju urnika najbolj izkaže model GPT-5. Končni izdelek je spletna aplikacija, ki vsakemu članu osebja omogoča izbiro preference dežurstva po dnevih v mesecu in pri mesečnem razporedu poskusi izpolniti vse zahteve.

• Jezik: Slovenski jezik
• Ključne besede: aplikacija, urnik, umetna inteligenca
• Vrsta gradiva: Diplomsko delo/naloga
• Organizacija: FRI - Fakulteta za računalništvo in informatiko
• Leto izida: 2026
• PID: 20.500.12556/RUL-178474
• Datum objave v RUL: 28.01.2026

Sekundarni jezik

• Jezik: Angleški jezik
• Naslov: Using artificial intelligence for schedule generation in health institutions

Izvleček

The contentment of employees is a very import aspect in working environment. Mandatory shifts, which we want to be in accordance with staff's wishes, also contribute to that. Because the current selection and assignment system in particular healthcare facility does not reach expectations, we decided to develop a system, which will provide improvements in that department. Instead of writing our own algorithm, we have tried to use the capabilities of artificial intelligence in solving the problem of scheduling. We monitored closely the products of company OpenAI, specifically models GPT-3.5-turbo, GPT-4.1 and GPT-5 and chatbot ChatGPT.We tested the models with the same input data and evaluated them based on their output, where we checked the evenness of the workload and the consideration of doctors' preferences and rest days.The chatbot had a different assignment, because it had to prepare the code, which then generates the schedule.We assessed the suitability of the code with the same input as the models.We found out that the model GPT-5 performs best in generating the schedule.The final products is a web application, that allows each staff member to select their shift preferences by day of the month and attempts to meet all requirements for created monthly schedule.

• Ključne besede: application, schedule, artificial intelligence