Naslov
Time-to-Event Prediction: Current Trends in Machine Learning

Autori
Stresec, Ivan

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, diplomski rad, diplomski

Fakultet
Fakultet elektrotehnike i računarstva

Mjesto
Zagreb

Datum
28.06

Godina
2022

Stranica
158

Mentor
Bojana Dalbelo Bašić

Ključne riječi
time-to-event ; censored data ; time-to-event prediction ; machine learning ; time-to-event analysis ; survival analysis ; reliability analysis

Sažetak
Time-to-event data is concerned with the time that passes from some point in time until an observed event happens. It is most commonly encountered in medicine (survival analysis) and engineering (reliability analysis) but is also present in other fields such as finance and sociology. The primary feature of time-to-event data is the occurrence of censorship, a mechanism that produces incomplete time data and prevents the use of standard regression methods. Important methods of time-to-event analysis, the branch of statistics concerned with time-to-event data, are discussed. Currently used machine learning models adapted for time-to-event prediction are described and examined in some detail, as are the relevant performance measures. Models are compared and analyzed in terms of performance on two datasets, one of which is concerned with medical data, the traditional domain of time-to-event prediction.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo

POVEZANOST RADA