RAZVOJ INFORMACIJSKE REŠITVE ZA OPTIMIZACIJO UPRAVLJANJA SISTEMA SOUPORABE KOLES

Božnar, Valerija (author), Kukar, Matjaž (mentor)

Sistemi souporabe koles so ključni del trajnostne mobilnosti, saj ponujajo učinkovito in okolju prijazno prevozno alternativo na krajših razdaljah ter izboljšujejo povezljivost z drugimi oblikami javnega prevoza. Kljub številnim prednostim se upravljavci pogosto soočajo z neravnovesji v razpoložljivosti koles, kar pomeni, da so nekatere postaje prepolne, druge pa prazne. To zmanjšuje dostopnost, slabša uporabniško izkušnjo in povečuje stroške upravljanja. Diplomsko delo se osredotoča na razvoj napovednega modela in administrativne platforme za podporo optimizaciji razpoložljivosti koles. Metodologija raziskave temelji na metodologiji CRISP-DM, ki vključuje faze razumevanja problema, zbiranja in priprave podatkov, izbire in umerjanja napovednih modelov, njihovega vrednotenja in implementacije. Analiza temelji na zgodovinskih podatkih izposoj koles ter dodatnih dejavnikih, kot so vremenske razmere, koledarske značilnosti in lokacijske posebnosti postaj. Uporabljene so kvantitativne metode in tehnike strojnega učenja za napovedovanje povpraševanja. Rezultati pokažejo, da vključitev vremenskih, koledarskih in lokacijskih spremenljivk bistveno izboljša natančnost napovedi potreb po prerazporeditvi koles. Razvita administrativna platforma omogoča enostaven dostop do podatkov, vizualizacijo zasedenosti postaj in generiranje priporočil za prerazporeditev. Uporabnost rezultatov se kaže v pomembnem prispevku k področju trajnostne mobilnosti in podatkovno podprtega upravljanja javnih storitev. Razvita rešitev omogoča upravljavcem sistemov souporabe koles učinkovitejše načrtovanje prerazporeditve, zmanjšanje operativnih stroškov in izboljšanje uporabniške izkušnje. Omejitev raziskave je uporaba podatkov iz omejenega časovnega obdobja, kar odpira priložnost za nadaljnje delo z vključitvijo daljših časovnih serij ter mehanizmov sprotnega učenja modela. Na družbeni ravni rešitev spodbuja trajnostno mobilnost, zmanjšuje prometne obremenitve in emisije ter prispeva k bolj dostopnemu in povezanemu urbanemu okolju.

• Language: Slovenian
• Keywords: javni sistem souporabe koles, upravljanje sistema, optimizacija, dejavniki izposoje in vrnitve, metodologija CRISP-DM, napovedni model, administrativna platforma
• Work type: Bachelor thesis/paper
• Organization: FU - Faculty of Administration
• Year: 2026
• PID: 20.500.12556/RUL-181351
• Publication date in RUL: 02.04.2026

Secondary language

• Language: English
• Title: DEVELOPMENT OF A SOLUTION FOR OPTIMIZING THE MANAGEMENT OF A BIKE-SHARING SYSTEM

Abstract

Bike-sharing systems are a key component of sustainable mobility, as they provide an efficient and environmentally friendly transport alternative for short distances and improve connectivity with other forms of public transport. Despite their many advantages, operators often face imbalances in bike availability, meaning that some stations are overcrowded while others are empty. This reduces accessibility, worsens the user experience, and increases operational costs. In this thesis, we therefore focused on the development of a predictive model and an administrative platform to support the optimization of bike availability. The research methodology was based on the CRISP-DM methodology, which includes the phases of problem understanding, data collection and preparation, selection and tuning of predictive models, their evaluation, and implementation. The analysis was based on historical bike rental data as well as additional factors such as weather conditions, calendar features, and station location characteristics. We applied quantitative methods and machine learning techniques to forecast demand. The results showed that incorporating weather, calendar, and location variables significantly improves the accuracy of predictions regarding bike redistribution needs. The developed administrative platform enables easy data access, visualization of station occupancy, and automatic generation of redistribution recommendations. The usefulness of the results is reflected in a significant contribution to the field of sustainable mobility and data-driven management of public services. The developed solution enables operators of bike-sharing systems to plan redistribution more efficiently, reduce operational costs, and improve the user experience. A limitation of the research is the use of data from a limited time, which opens opportunities for future work involving longer time series and mechanisms for real-time model learning. At the societal level, the solution promotes sustainable mobility, reduces traffic congestion and emissions, and contributes to a more accessible and connected urban environment.

• Keywords: public bike-sharing system, system management, optimization, rental and return factors, CRISP-DM methodology, predictive model, administrative platform