Darin Goldstein

Title: Darin Goldstein

Main Research Question: How can artificial intelligence be used to select the best model for predicting the probability of an individual having a particular disease, given a large population with various traits?

Methodology: The researchers applied Adaptive Simulated Annealing (ASA), an optimization algorithm, to a large data set containing information on a population's characteristics. They compared ASA's performance to traditional forward and backward regression methods. The goal was to determine which method performed best on larger, more complex data sets.

Results: The study found that ASA outperformed the traditional methods on computer-simulated data sets with varying sizes (100,000, 500,000, and 1,000,000). The ASA method consistently produced a model with a Cp statistic, which reflects bias and variance, that was close to optimal. In contrast, forward and backward regression occasionally did not produce the best results.

Implications: These findings suggest that ASA is a promising approach for model selection in large data sets, particularly in epidemiological settings. The method's ability to select the most informative model without overfitting the data could lead to more accurate predictions and better public health interventions.

In summary, Darin Goldstein's research demonstrates the potential of using Adaptive Simulated Annealing for model selection in large data sets, particularly in epidemiological settings. The study's findings suggest that ASA can produce more accurate models than traditional methods, leading to better predictions and interventions.

Link to Article: https://arxiv.org/abs/0310005v1 Authors: arXiv ID: 0310005v1