diplomsko delo
Domen Kavran (Author), Niko Lukač (Mentor)

Abstract

V diplomskem delu opišemo algoritem segmentacije časovnih vrst in postopek priprave vektorjev značilnic segmentov za učenje in testiranje klasifikacijskih modelov za zaznavo dogodkov. Segmentacijo časovnih vrst izvedemo z algoritmom drsečega okna, kjer za merilo razdalje med vrednostmi uporabimo algoritem dinamičnega časovnega sledenja. Pripravo vektorjev značilnic segmentov začnemo z definiranjem slovarja lokalnih podsegmentov. Slovar je pridobljen z gručenjem K-povprečij. Vsak segment predstavimo z normaliziranim histogramom pojavitev lokalnih podsegmentov na podlagi slovarja. Za učenje klasifikacijskih modelov uporabimo algoritme strojnega učenja, ki se razlikujejo v računski zahtevnosti in doseženi natančnosti, na katero vplivajo tudi izbrani parametri segmentacije in velikost slovarja.

Keywords

klasifikacija;časovna vrsta;strojno učenje;segmentacija;diplomske naloge;

Data

Language: Slovenian
Year of publishing:
Typology: 2.11 - Undergraduate Thesis
Organization: UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher: D. Kavran
UDC: 004.5:004.852(043.2)
COBISS: 21746198 Link will open in a new window
Views: 1537
Downloads: 169
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Classification of events in time series data using machine learning
Secondary abstract: In this thesis, an algorithm of time series segmentation and procedure of preparing segments feature vectors for training and testing classification models are presented, in order to detect time series events. Sliding window algorithm with dynamic time warping as distance measure is used for time series segmentation. Creating segments feature vectors starts with defining a dictionary of local subsegments. Dictionary is created with K-means clustering. Each segment is described with normalized histogram of local subsegment occurances based on dictionary. Machine learning algorithms, used for training classification models, differ in computation complexity and achieved accuracy. Achieved accuracy depends on the selected segmentation parameters and dictionary.
Secondary keywords: classification;time series;machine learning;segmentation;
URN: URN:SI:UM:
Type (COBISS): Bachelor thesis/paper
Thesis comment: Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije
Pages: X, 29 f.
ID: 10949570