S porastom količine podatkov se nam možnosti za njihovo obdelavo in odkrivanje novih znanj širijo. Z napredkom tehnologije in metod obdelave podatkov se tako hitrost obdelave kot kvaliteta dobljenih informacij izboljšuje, a hkrati starejše, vendar ne nujno slabše metode, padejo v pozabo. Z združitvijo metod na enem mestu in njihovo medsebojno primerjavo dosežemo razširljiv sistem za primerjavo in izbor najustreznejše metode za problem, ki ga obravnavamo.
V našem delu smo pregledali nekatere obstoječe metode za odstranjevanje trendov iz ritmičnih podatkov, jih implementirali in jih primerjali med sabo.
V primerjavi smo ugotovili, da nekatere metode precej bolj ustrezajo nekaterim tipom podatkov. V našem delu smo primerjali izbran nabor metod, sistem pa je razširljiv, tako da je dodajanje novih metod preprosto. Enako velja za poganjanje metod na novih podatkih, saj metode sprejemajo precej preprost format podatkov. Implementacija metod je na voljo prihodnjim uporabnikom na sledečem repozitoriju:
https://github.com/mmoskon/CosinorPy.
| Language: | English |
|---|
| Title: | The overview, implementation, and evaluation of rhythmic data preprocessing methods |
|---|
| Abstract: |
|---|
With the increase in the amount of data our options for processing that data and discovering new knowledge keep expanding. The advancement of technology and data processing methods leads to increased processing speeds and greater quality of information obtained from the data while older but not necessarily worse methods fall out of memory. With the aggregation of these methods in one place and their comparison, we create an expandable system for the comparison and selection of the most fitting method for the problem we are handling.
In our work, we made an overview of a selected set of existing methods for detrending rhythmic data, implemented them, and compared them with one another.
In our comparison, we concluded that some methods are a much better fit for some types of data. In our work, we compared many methods and the system is expandable, so adding new methods is simple. The same goes for running the methods on new data, as the methods accept a very simple data format. The implementation of the used methods is available for future users on the following repository: https://github.com/mmoskon/CosinorPy.
|
| Keywords: | rhythmic data, data preprocessing, preprocessing methods, data analysis |
|---|
