Ensemble Learners for Boosted Explanatory Predictive Model of Lung Cancer Outcomes

Principal Investigator

Name
Ahmed Hassoon

Degrees
M.D., M.P.H., P.M.P.

Institution
Johns Hopkins Bloomberg School of Public Health

Position Title
Research Associate

Email
ahassoo1@jhu.edu

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-407

Initial CDAS Request Approval
Apr 27, 2018

Title
Ensemble Learners for Boosted Explanatory Predictive Model of Lung Cancer Outcomes

Summary
I would like to use the NLST data to build a sent of ensemble weak learners to predict diagnostic errors and health related outcomes. The weak learners will work by distributing the predictive task among weak learners, and learn from the mistakes of prior learners to enhance the prediction of the following learner. All orchestrated (musical ensemble) to boost prediction. I will focus on sequential explanatory prediction to transition from one state to another state to minimize noise.

Parallel to that, I will use sequential data mining to discover hidden patterns that may inform our predictive model.

Aims

1- To test if ensemble of weak learners will improve prediction of lung cancer related outcomes.
2- To build a predictive decision model at different time point for adaptive prediction.
3- To conduct sequential data mining to discover hidden patterns, clustered patterns, and clusters patterns that have relation to the outcome(s) of interest.
4- Comparing input data to their input image representation in predicting lung cancer among the NLST participants.
5- Assess the proportion of participants in the National Lung Cancer Screening Trial (NLST) who experienced diagnostic uncertainty after a positive lung cancer screening.

Outcome of interests: Diagnostic Accuracy, Diagnostic Errors, Diagnostic Staging, Death, Other Health related outcomes.

Collaborators

None. I welcome any collaborator !