iSurvive: An Interpretable, Event-time Prediction Model for mHealth

Authors: Walter H. Dempsey, Alexander Moreno, Christy K. Scott, Michael L. Dennis, David H. Gustafson, Susan A. Murphy, James M. Rehg

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We show promising results both in simulation and on a real-world m Health recovery support services dataset from individuals with substance use disorders. and For each experiment that follows, we perform a crossvalidation based assessment of prediction accuracy. and Figure 2. Cross-validated complete Brier score on recovery support services study for several discriminative models and i Survive
Researcher Affiliation Collaboration 1 University of Michigan 2 Georgia Institute of Technology 3 Lighthouse Institute 4 University of Wisconsin Madison.
Pseudocode Yes Algorithm 2 Event prediction algorithm
Open Source Code Yes Our publicly-released software (http://cbi.gatech.edu/Survival-HMM) will enable the m Health and data science communities to benefit from these new modeling capabilities.
Open Datasets No We analyze a set of recovery support studies on individuals with substance use disorders (SUDs). In particular, we analyze two pilot studies a 5-week study of adults (N = 23) and a 6-week study of adolescents (N = 29)... The paper describes a specific collected dataset for their case study but does not provide access information (link, DOI, citation) to make it publicly available.
Dataset Splits Yes For each experiment that follows, we perform a cross-validation based assessment of prediction accuracy. We randomly partition the N participants into groups of size K. Suppose N/K = M and we label each partition uniquely m = 1, . . . , M
Hardware Specification No The paper mentions 'smartphones' used for data collection but does not specify any hardware (e.g., GPU, CPU models, or cloud instances) used to run the experiments or train the models.
Software Dependencies No The paper discusses algorithms and models (e.g., GLM, CT-HMM, EM algorithm, Fisher scoring) but does not specify any software libraries or dependencies with version numbers.
Experiment Setup No The paper describes the model architecture, estimation methods (EM), and initialization strategies, but it does not provide specific hyperparameter values (e.g., learning rate, batch size, number of epochs) or system-level training settings.