How to diagnose memory leaks in Java real-time processing under peak load?

[Penelope.J]

I'm working on a Java application that processes large datasets in real-time, and we're hitting persistent OutOfMemoryError exceptions during peak loads, even after increasing the heap size. I suspect the issue is less about the total memory and more about inefficient object creation and retention in our processing loops. I'm looking for practical strategies or tools to better analyze our heap dumps and identify memory leaks, specifically around collections and cached objects. What are the most effective profiling techniques or JVM flags you've used to optimize Java memory management in a high-throughput environment?