Details

Primerjava algoritmov strojnega učenja za napovedovanje bodočih stanj medicinske naprave
ID KEKIĆ, ADMIR (Author), ID Kos, Andrej (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (3,88 MB)
MD5: 45607C543CD93F3C26E8380A89D7E38D

Abstract
V diplomski nalogi sem analiziral možnosti napovedovanja senzornih podatkov medicinskega stroja WD 290, ki ga proizvaja podjetje Belimed. Namen napovedovanja senzornih podatkov je omogočiti zgodnje zaznavanje morebitnih nepravilnosti v delovanju stroja, ter s tem preprečiti izpade ali okvare. Napovedovanje tako služi kot orodje za oceno prihodnjega stanja stroja na podlagi trendov v senzornih podatkih. Takšen pristop lahko služi kot osnova za vzpostavitev prediktivnega servisa, ki omogoča pravočasno vzdrževanje in optimizacijo delovanja naprav. Z uporabo različnih algoritmov strojnega učenja (linearna in polinomska regresija, odločitveno drevo ter metod k-najbližjih sosedov - KNN) sem primerjal njihovo učinkovitost pri napovedovanju vrednosti na podlagi zbranih podatkov. Podatke sem pridobil s pomočjo programske opreme SmartHub ter jih obdelal z uporabo programskega jezika Python. Izvedel sem predobdelavo podatkov, učenje modelov ter njihovo vrednotenje z metrikama koeficientom determinacije in povprečno kvadratno napako. Ugotovil sem, da sta se metodi odločitvenega drevesa in KNN izkazali za najbolj natančni, medtem ko sta regresijska modela pokazala slabšo zmogljivost zaradi kompleksnosti podatkovnega nabora.

Language:Slovenian
Keywords:Napovedovanje, strojno učenje, regresija, odločitveno drevo, k-najbližjih sosedov, obdelava podatkov, analiza podatkov.
Work type:Bachelor thesis/paper
Typology:2.11 - Undergraduate Thesis
Organization:FE - Faculty of Electrical Engineering
Year:2025
PID:20.500.12556/RUL-175830 This link opens in a new window
COBISS.SI-ID:266955779 This link opens in a new window
Publication date in RUL:10.11.2025
Views:475
Downloads:134
Metadata:XML DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:English
Title:Comparison of machine learning algorithms for predicting future states of a medical device
Abstract:
In this thesis, I analyzed the possibilities of predicting sensor data from the WD 290 medical machine, manufactured by the company Belimed. The purpose of sensor data prediction is to enable early detection of potential irregularities in the machine’s operation, thereby preventing breakdowns or failures. Prediction thus serves as a tool for assessing the future condition of a machine based on trends in sensor data. Such an approach can serve as a basis for establishing a predictive service, which enables timely maintenance and optimization of device performance. Using various machine learning algorithms (linear and polynomial regression, decision tree, and k-nearest neighbors - KNN), I compared their effectiveness in forecasting values based on the collected data. The data was acquired using the SmartHub software and processed with the Python programming language. I performed data preprocessing, model training, and evaluation using the metrics of coefficient of determination and mean squared error. The result showed that the decision tree and KNN methods were the most accurate, while the regression models performed poorly due to the complexity of the dataset.

Keywords:Predicting, machine learning, regression, decision tree, k-nearest neighbors, data processing, data, data analysis.

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back