It tracks the CPU utilization for the database belonging to the application, so you could say it is caused by the Mendix application.
From the documentation: First actions to take: Inspect the trends for Application node CPU usage combined with Application node disk throughput and Application node load for anomalies and correlate those with application behavior.
What you need to figure out is what is running during those times. Are there any scheduled events that can be linked to these timestamps that are processing large amounts of data or incoming API calls that put a load on the application and database.
Try to find what is running during that time and if the load can be reduced or spread over time.