capstone-project

Title

Machine Learning Approach to Finding Similar Geological Samples in EarthChem-PetDB Database

Reproducibility

• clone the repository
• on terminal: cd capstone-project
• create the conda environment from the yml file: conda env create -f environment.yml
• activate the environment: conda activate final-project
• preprocess the available data and create useful datasets: python createDataset.py
• create label distribution plots: python visuals.py
• run experiments with logistic regression model: python logisticRegression.py
• run experiments with xgboost model: python xgb.py
• run experiments with neural networks model: python nn.py
• run experiments with all models but with leave one out cross validation: python leaveOneOut.py, however this takes a while due to the nature of LOOCV method
• run the model explainer to see which features matter how much: python explainer.py
• run the inference to get top n similar samples given the new samples: python inference.py, remove the break in the loop if you want the results for all new samples