Loading repository data…
UrbsLab / repository
An rigorous, well documented machine learning analysis pipeline for binary classification datasets assembled as a Jupyter Notebook. Includes exploratory analysis, data processing, feature processing, ML modeling (9 algorithms, including the original ExSTraCS algorithm) with hyperparameter sweeps, visualizations, and statistical analysis. A comprehensive starting point to adapt to your own dataset an as an example of how to integrate a non-scikit-learn ML algorithm into a comparative pipeline.
A transparent discovery signal based on current public GitHub metadata.
This score does not audit code, security, maintainers, documentation quality, or suitability. Verify the repository and its current documentation before adoption.
This respository includes the code (including a Jupyter Notebook) to run the machine learning analysis pipeline (for binary classification) that pairs with our paper titled "A Rigorous Machine Learning Analysis Pipeline for Biomedical Binary Classification: Application in Pancreatic Cancer Nested Case-control Studies with Implications for Bias Assessments". This notebook presents an example of a 'rigorous' and well annotated machine learning (ML) analysis pipeline that could be reasonablly applied to various supervised learning classification tasks, but was developed here specifically for biomedical data mining/modeling. This pipeline could be used as follows:
This code represents our first ML analysis pipeline implementation as described in the paper above. It includes the ExSTraCS machine learning modeler which is a stand alone python implementation (outside of scikit-learn). More recently we have implemented scikit-learn compatible versions of ExSTraCS and two other LCS algorithms (i.e. XCS, and eLCS). We have since implemented an updated/improved version of this ML analysis pipeline to include those three new scikit-learn compatible LCS algorithms along with nearest neighbors and gradient boosting classifiers. For users looking to replicate the pipeline in the above paper, use this present repository. For those looking for the most up-to-date pipeline including 5 additional ML modeling algorithms we recommend using the very similar pipeline available at: https://github.com/UrbsLab/scikit_ML_Pipeline_Binary_Notebook
In order to run this pipeline as a Jupyter Notebook you must have the proper environment set up on your computer. Python 3 as well as a number of Python packages are required. Most of these requirements are satisfied by installing the most recent version of anaconda (). We used Anaconda3 with python version 3.7.7 during this pipeline development. In addition to the packages included in anaconda, the following packages will need to be installed separately (or possibly updated, if you have an older version installed). We recommend installing them within the 'anaconda prompt' that installs with anaconda:
Selected from shared topics, language and repository description—not editorial ratings.
A Python project dedicated to predicting customer churn within the telecommunications sector. Leveraging a dataset of customer interactions and demographics, this project dives deep into rigorous Exploratory Data Analysis to uncover underlying patterns and behaviors factors contributing to customer churn, as well as building a classifier model.
27/100 healthAdditionally, while currently commented out in the file (modeling_methods.py) if you want the optuna hypterparameter sweep figures to appear within the jupyter notebook (via the command 'fig.show()' ) you will need to run the following installation commands. This should only be required if you edit the python file to uncomment this line for any or all of the ML modeling algorithms.
Lastly, in order to include the stand-alone algorithm 'ExSTraCS' we needed to call this from the command line within this Jupyter Notebook. As a result, the part of this notebook running ExSTraCS will only run properly if the path to the working directory used to run this notebook includes no spaces. In other words if your path includes a folder called 'My Folder' vs. 'My_Folder' you will likely get a run error for ExSTraCS (at least on a Windows machine). Thus, make sure to check that wherever you are running this notebook from, that the entire path to the working directory does note include any spaces.
This notebook loads a single dataset to be run through the entire pipeline. Here we summarize the requirements for this dataset:
Included in this repository is the following:
The user can turn on/off the option to filter out the lowest scoring features in the data (i.e. to conduct not just feature importance evaluation but feature selection)
Feature importance evaluation and feature selection are conducted within each respective CV training partition
The pipeline reports feature importance estimates via two feature selection algorithms:
When selected by the user, feature selection conservatively keeps any feature identified as 'potentially relevant' (i.e. score > 0) by either algorithm
Since MultiSURF scales quadratically with the number of training instances, there is an option to utilize a random subset of instances when running this algorithm to save computational time.