Relyon – Preprocessing Optimization Interface

Postado em 17/Oct/2024

Relyon is a free standalone interface built in Matlab for the optimization of multivariate models using PLS regression. The interface was developed by Wilson Cardoso and Jussara Roque at MCDALab under the supervision of Prof. Reinaldo Teófilo. The interface is capable of importing .mat or .xlsx files, automatically optimizing preprocessing steps, providing the option to automatically split data into training and testing sets, selecting variables using the OPS method, building PLS calibration models, and making predictions for new unknown samples.

The interface offers 21 preprocessing options that can be executed individually or in combination. The results are saved in an Excel spreadsheet, allowing the user to decide on the best strategy. For the Ordered Predictors Selection (OPS), four methods are available (AutoOPS, FeedOPS, AutoiOPS, and FeediOPS). Model parameters are presented in a table within the interface, where the user can choose the best model and save it in .dat format for later use in predicting new samples.

The user also has the ability to save .xlsx files with desired data for creating graphs, as well as saving model parameters and selected variables. Relyon is a comprehensive software for the creation and optimization of PLS models.

The paper “Combinatorial Order Pre-processing Search (COPS): A New Pre-processing Strategy for Large-Scale Interpretable Data Analysis in Process Analytical Technologies” discussing preprocessing optimization was published in Computers & Chemical Engineering Journal and can be accessed at https://doi.org/10.1016/j.compchemeng.2024.108892.

The Relyon Interface can be downloaded at https://github.com/wjcardoso/relyon.