JANOS is an integrated predictive and prescriptive modeling framework. It seamlessly integrates two streams of analytics, discrete optimization and predictive modeling, enabling practitioners to make optimal data-driven decisions.
JANOS allows specifying the prescriptive model using the standard optimization modeling elements such as constraints and variables. The key novelty lies in providing modeling constructs that allow for the specification of commonly used predictive models and their features as constraints and variables in the optimization model. JANOS considers two sets of decision variables: regular and predicted. The relationship between any pairs of regular and predicted variables can be specified by users as pre-trained predictive models.
JANOS currently supports linear regression, logistic regression, and neural network with a rectified linear activation function, but we plan to expand on this set in the future.