Skip to content
thesis project. Building predictive modeling using statistical methods, machine learning and deep learning techniques
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
experiments
src
Gender_CV_STRATA-github.ipynb
Kfold CV - Check error across gender_osteo.ipynb
README.md
Race-LR.ipynb
current_results.md
data_cleaning.py
redo decile calcium and sodium.ipynb
train_models.py
write function to train model_backup.ipynb

README.md

Machine-Learning

thesis project. Building predictive modeling using statistical methods, machine learning and deep learning techniques

There are 2 datasets, both are case-control study:

  1. osteoporosis, case-control for osteoporosis
  2. bone-fracture, case-control for bone-fracture

Notice this is case-control study, i.e. CV must based on STRATA instead of doing normally

Techniques to be applied:

  • Logistic Regression
  • SVM (non-linear)
  • Random Forest (non-linear)
  • ANN

Update results:

  • because this is case-control study, there should be no bias in subgroups/demographic, i.e. AUC should be the same across subgroups. But current results do not support that!!! Train using LR and RF to track non-linear relationship, both LR and RF not the same either train subgroup together or train them separately
You can’t perform that action at this time.