| Title |
Functional data classification by depth measures. case study of telecommunication data / |
| Translation of Title |
Funkcinių duomenų klasifikavimas pagal gylį. Telekomunikacinių duomenų atvejo analizė. |
| Authors |
Balužis, Raigardas |
| Full Text |
|
| Pages |
44 |
| Keywords [eng] |
Functional Data Analysis, Classification, Functional depth, FANOVA, Telecommunications |
| Abstract [eng] |
The thesis proposes a Functional Data Analysis approach to tackle real telecommunication industry problems. The study includes the applications of monotone smoothing, functional outliers detection and FANOVA methods. However, the main focus is concentrated towards proposing an efficient supervised functional classification method, that could be used in the telecommunication industry. Thus, multiple depth-based classification techniques are used to identify upcoming customer action based on an individual's behaviour when consuming mobile data. A comprehensive comparison of functional data classifiers is provided, and the influence of three main parameters is explored. The comparative study results demonstrate the relatively high accuracy of Maximum depth classifiers and the impact of different depth measures towards the classification results. Based on the accuracy, the Maximum depth classifier using the Fraiman and Muniz depth method emerges as the most efficient classifier in the case study. |
| Dissertation Institution |
Vilniaus universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2021 |
