clarencew0083/recSystem: Meta Learning Recommendation System

This package extracts meta features from a dataset to recommend what machine learning algorithm will perform the best without running all the implemented machine learning algorithms. The final data product will be a GUI that will ingest a data set in order to predict a response using classification or regression and display to the user the ranking of the machine learning algorithms on the target data set. The current selection of algorithms is limited to linear regression, ridge regression and support vector regression for continuous responses and k nearest neighbors classifier, naive bayes classifier and support vector classifier for binary output. Additionally, evaluation metrics for each algorithm include normalized root mean square error for continuous responses and precision/recall for binary output. The meta learner utilizes dimension reduction via Principal Component Analysis and use support vector regression with a gaussian kernel to predict the evaluation metric. The final data product will be a R shiny application as the front end with python integration to call the meta learn recommendation system functions. Python 3.7 is the required version of python. Numpy 1.17.4, pandas 0.25.1, and sci-kit learn 0.21.3 are the required python packages. While not necessary end users should be familiar with the implemented machine learning algorithms, if they plan to implement the one recommended algorithms. There are no security concerns with this product and there are no design constraints.

install.packages("remotes")
remotes::install_github("clarencew0083/recSystem")