The problem of tracking file system events and the sequences in which they occur is quite similar to the text compression problem of tracking strings of characters to model their frequency distributions. The use of data compression modeling techniques for prefetching in operating systems was first investigated Vitter, Krishnan and Curewitz [19,4]. It was their work that inspired us to adapt PPM's modeling techniques, to track file system events instead of characters of an alphabet. Using the information tracked by this model we chose a method based on a likelihood threshold to select the events to prefetch.