Systems, methods, and computer program products for automating the management of a tiered storage system, which improves the service life and costs of the flash-based storage devices that comprise the storage system. The systems, methods and program products utilize time series learning models to predict the write frequency of each data block for the next cycle of time. Using the write frequency predictions, management of the tiered flash-based storage system can automatically organize and migrate stored data based on storage costs, migration costs and frequency in which the data is accessed. Data that is more frequently accessed and updated is migrated to high-end flash-based storage devices, which are able to endure a greater number of program/erase cycles, while less frequently accessed data is migrated to less expensive, low-end flash-based storage devices, that have a lower maximum number program/erase cycles.