Abstract: A method and system for predicting incidents in an application environment are disclosed. The method includes receiving, by the at least one processor, a set of data from a plurality of data sources associated with the application environment. The method further includes unifying, by the at least one processor, the received set of data into a unified set of metrics. The method further includes analyzing, by the at least one processor using a trained model, the unified set of metrics to identify a set of anomalies indicative of a potential incident. The method further includes validating, by the at least one processor, the identified set of anomalies to confirm the potential incident. Thereafter, the method includes generating, by the at least one processor, an alert based on the confirmed potential incident.