BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Justine Remiat — Random forests using longitudinal predictors
DTSTART:20250417T110000Z
DTEND:20250417T120000Z
DTSTAMP:20260423T064700Z
UID:indico-event-14242@indico.math.cnrs.fr
CONTACT:boris.hejblum@u-bordeaux.fr
DESCRIPTION:Speakers: Justine Remiat\n\nSpeaker: Justine Remiat from Bord
 eaux Population healthTitle: Random forests using longitudinal predictors\
 nThis seminar will be in English\nAbstract: Random Forests (Breiman\, 2001
 ) are an effective predictive tool\, particularly in high-dimensional set
 tings. However\, they are not well-suited for longitudinal data collected 
 over time. To address this limitation\, Fréchet Random Forests (Capitaine
  et al. 2020) were proposed. They can handle any type of data within a met
 ric space by using a distance tailored to each data type (e.g.\, images\, 
 trajectories). This work aimed to implement the Fréchet Random Forest for
  trajectory data\, fully exploiting the flexibility of the generalized dis
 crete Fréchet distance\; and evaluate the performance of the Fréchet Ran
 dom Forest in predicting a continuous outcome using longitudinal inputs. T
 he Generalized Discrete Fréchet Distance depends on a time-shifting param
 eter\, called timescale\, which modifies its behavior. We proposed two im
 plementations: the time-scale defined as an hyper parameter or the time-sc
 ale randomly drawn at each tree node to explore all time sensitivity behav
 iors. A simulation study has been conducted to illustrate the flexibility 
 of the Fréchet Random Forest to capture different scenarios of associatio
 n: (i) time-sensitive association (ii) shape-sensitive association and (ii
 i) a mix of both. We then apply the method to data from a population-based
  cohort to predict the risk of dementia from clinical marker trajectories.
  The simulations illustrated the flexibility of the Fréchet Random Forest
 s to adapt to different types of associations with the timescale tuning. T
 he Fréchet Random Forests also demonstrated better predictive performance
  (MSE) across all three scenarios compared to classical Random Forests wit
 h pre-determined features. On the application data\, the Fréchet forests 
 outperformed classical forests\, even with more irregular and sparse data\
 , while similarly identifying predictive markers. Thanks to its tunable ti
 mescale parameter that can adapt to different structures of association\, 
 the Fréchet Random Forest constitutes a flexible tool for prediction base
 d on longitudinal data.\n \nCalendar subscription link for the complete s
 eminar series:https://indico.math.cnrs.fr/category/711/events.ics\nProgram
  of the Biostatistics seminars:https://indico.math.cnrs.fr/category/711/\n
 Subscribe to the seminar mailing list:https://diff.u-bordeaux.fr/sympa/sub
 scribe/seminaire.biostat.bph\nFormer e-seminars on our YouTube channel (mo
 stly in French): https://www.youtube.com/channel/UCURp-hEQL7k23UzGfqgEurA/
 videos\n \nBiostatistics seminar series from the Department of Public Hea
 lth from the University of Bordeaux and the Bordeaux Population Health UMR
  1219 research center\n \n\nhttps://indico.math.cnrs.fr/event/14242/
LOCATION:Amphi Louis (ISPED)
URL:https://indico.math.cnrs.fr/event/14242/
END:VEVENT
END:VCALENDAR