SEQUENCE PREDICTION FOR PRINT/SCAN ISSUES USING BAYESIAN NETWORKS

Machine Learning algorithms are empowering a lot of software applications today. We address a business need where the Machine Learning model can help to resolve customer issues which arise during printing and scanning on devices or installing a software driver. The model is based on Bayesian Network diagnoses the problem and subsequently resolves it by suggesting a sequence of steps which increase the probability of fix. This is applied in the context of an installer or driver or a diagnostics tool (referred to as *component*) to arrive at a resolution. The model factors in the current state of the print system which comprises of Operating System, Printer model, localization etc. The resolution can be a static sequence which provides a sequence of steps based upon the initial state of the customer or it can be a dynamic sequence wherein user provides feedback against each action suggested by the model and the model takes that feedback into account and suggests the next step. We have a prototype for this model and are applying it as part of Machine Learning service of Print Scan Doctor (PSDr) and are evaluating the possibility of other applications