Documentation: https://raamana.github.io/neuropredict/
News
- As of
v0.6, neuropredict now supports regression applications i.e. predicting continuous targets (in addition to categorical classes), as well as allow you to regress out covariates / confounds within the nested-CV (following all the best practices). Utilizing this feature requires the input datasets be specified in thepyradigmdata structures: code @ https://github.com/raamana/pyradigm, docs @ https://raamana.github.io/pyradigm/. Check the changelog below for more details.
Older news
neuropredictcan handle missing data now (that are encoded withnumpy.NaN). This is done respecting the cross-validation splits without any data leakage.
Overview
On a high level,
On a more detailed level,
- Docs: https://raamana.github.io/neuropredict/
- Contributors most welcome: check ideas and the following guidelines. Thanks.
Long term goals
neuropredict, the tool, is part of a broader initiative described below to develop easy, comprehensive and standardized predictive analysis:
Citation
If neuropredict helped you in your research in one way or another, please consider citing one or more of the following, which were essential building blocks of neuropredict: – Pradeep Reddy Raamana. (2017, November 18). neuropredict: easy machine learning and standardized predictive analysis of biomarkers (Version 0.4.5). Zenodo. http://doi.org/10.5281/zenodo.1058993 – Raamana et al, (2017), Python class defining a machine learning dataset ensuring key-based correspondence and maintaining integrity, Journal of Open Source Software, 2(17), 382, doi:10.21105/joss.00382
Change Log – version 0.6
- Major feature: Ability to predict continuous variables (regression)
- Major feature: Ability to handle confounds (regress them out, augmenting etc)
- Redesigned the internal structure for easier extensibility
- New
CVResultsclass for easier management of a wealth of outputs generated in the Classification and Regression workflows - API access is refreshed and easier
Change Log – version 0.5.2
- Imputation of missing values
- Additional classifiers such as
XGBoost, Decision Trees - Better internal code structure
- Lot more tests
- More precise tests, as we vary number of classes wildly in test suites
- several bug fixes and enhancements
- More cmd line options such as
--print_optionsfrom a previous run
