Logit enables automatic detection of functional dependencies in measured data by means of training and validation of logistic regression models.

Main features

• Data processing and normalization, advanced data imputation, data noise reduction and automated selection of the most important attributes from a large attribute set.
• Testing of created hypotheses, saving the trained models, classification of unclassified data, data and result visualization.
• Estimation of optimal number of samples to avoid logistic regression model overfitting based on generating samples from Guassian mixture model.

Example usage: Identification of differences in immunity parameters between patients using a tested drug and patients using placebo. A two-dimensional prediction model, which discriminates patients with 77% accuracy, has been identified on a set of patients (18 placebo, 35 drug).

Logit can run both locally and in server mode where computationally intensive parts of algorithm run on server.

Logit was developed in cooperation with the Cancer Research Institute of the Slovak Academy of Sciences.