Some rule generation algorithms lend themselves considerably better to parallelisation
than others. An early attempt to parallelise the TDIDT decision
tree induction algorithm is described in [5]. The Prism algorithm for generating
modular rules described in Chapter 11 is also one that lends itself well to
this approach. The PMCRI (Parallel Modular Classification Rule Induction)
framework [4, 6, 7] was developed by the German researcher Dr. Frederic Stahl
in association with the present author as a distributed version of Prism. In this
and the following section PMCRI will be used as a vehicle for explaining some
general principles, but the algorithm itself will not be described in detail here.
The account in these two sections draws heavily from [4]. Figures 13.5 to 13.7
are reproduced from [4] and Figures 13.4 and 13.8 are reproduced from [6] with
permission.