$ Predictive performance of Mixed-Effects Cox regression and Learning Neural Network model with application in agriculture $
Keywords: Censoring data, Mixed effect Cox model, Learning neural networks, Performance.
Abstract: $The Mixed effects Cox model and survival neural networks are commonly used to
predict the hazard rate of an interesting event. This paper compared the predictive
ability of the two techniques using simulation tools based on sample size and
the censoring rate of mixed effects data from agriculture. Thus, varying into
three sample sizes and five censoring rates, fifteen different datasets with 1000
replications were simulated based on the characteristics of a real data from Jatropha
Curcas L. seeds germination trial. Based on three performance metrics (Brier Score,
Integrate Brier Score, and Concordance index), the mixed effects Cox model better
predicted the survival outcomes than the Partial Logistic Artificial Neural Network
original on relatively low sample size (768 or 1536) and middle censoring rate (40
% or 50.65 %). Therefore, the dimension and censoring rate of the relevant dataset
must be considered when selecting one of the two models for analysis. $
Submission Category: Machine learning algorithms
Submission Number: 87
Loading