Prediction Diagnostics - Addressing data veracity in predicting batch processes

In today's IT world, batch systems empower several industry verticals for doing complex data-driven operations. Batch systems are defined as sequence of jobs that perform different tasks to complete various business operations. Any delay in the execution of these operations has a negative implication on the operations of the organization. As a result, industries across the globe are attempting to transform their batch operations from reactive to proactive. In this paper, we present an approach to improve predictions that can assist in this transformation. We propose to provide a guided diagnostic to identify causes of prediction errors and provide corrective recommendations. The concepts present in this paper are developed over our experience with several real-world case study that demonstrate the effectiveness of the proposed approach in improving predictions.