magistrsko delo
Povzetek
Namen naloge je izdelava matematičnih modelov napovedovanja za odločitve upravljanja, osnovane na inteligentnih, kvantitativnih analizah. Magistrsko delo obravnava področje napovedovanja števila interventnih dogodkov Gasilske brigade Maribor s pomočjo umetne inteligence in regresijskih modelov. Učne množice podatkov so bile pridobljene iz baz podatkov SPIN in ARSO, obdelane v programskem jeziku Python, modeli napovedovanja pa programirani v programskem paketu MATLAB. Cilj naloge je bil izdelava štirih regresijskih algoritmov, umetne nevronske mreže LSTM in NARX za napovedovanja dogodkov, njihove rezultate pa preko metrik ocenjevanja natančnosti medsebojno primerjati. Rezultati napovedovanja nekaterih učnih množic so bili zaradi majhnih korelacijskih povezav slabi, zato teh dogodkov nismo mogli napovedovati. Požarne intervencije in naravne nesreče so dale dovolj dobre rezultate korelacijskih analiz, zato so bile uporabljene v izgradnji nevronskih mrež. Glede na rezultate zbranih modelov menimo, da so nevronske mreže primernejše za napovedovanje interventnih dogodkov kot regresijski modeli.
Ključne besede
napovedovanje;umetna inteligenca;nevronske mreže;strojno učenje;regresija;magistrske naloge;
Podatki
Jezik: |
Slovenski jezik |
Leto izida: |
2020 |
Tipologija: |
2.09 - Magistrsko delo |
Organizacija: |
UM FS - Fakulteta za strojništvo |
Založnik: |
[R. Rutnik] |
UDK: |
004.8(043.2) |
COBISS: |
38183427
|
Št. ogledov: |
764 |
Št. prenosov: |
55 |
Ocena: |
0 (0 glasov) |
Metapodatki: |
|
Ostali podatki
Sekundarni jezik: |
Angleški jezik |
Sekundarni naslov: |
Interventions prediction using artificial intelligence |
Sekundarni povzetek: |
The purpose of the thesis is to create mathematical forecasting models for management decisions based on intelligent, quantitative analyzes. The master's thesis deals with the field of predicting the number of intervention events of the Maribor Fire Brigade with the help of artificial intelligence and regression models. Learning data sets were obtained from SPIN and ARSO databases, processed in the Python programming language, and prediction models were programmed in the MATLAB software package. The aim of the task was to develop four regression algorithms, an artificial neural network LSTM and NARX for predicting events, and to compare their results with each other through metrics for estimating accuracy. The prediction results of some learning sets were poor due to small correlations, so we could not predict these events. Fire interventions and natural disasters gave good enough results of correlation analyzes, so they were used in the construction of neural networks. Based on the results of the collected models, we believe that neural networks are more suitable for predicting intervention events than regression models. |
Sekundarne ključne besede: |
forecast;artificial intelligence;neural networks;machine learning;regression; |
Vrsta dela (COBISS): |
Magistrsko delo/naloga |
Komentar na gradivo: |
Univ. v Mariboru, Fak. za strojništvo, Proizvodne tehnologije in sistemi |
Strani: |
XII, 80 str. |
ID: |
12030293 |