An investigation into the effect of cluster-based preprocessing on software migration

The problem of language obsolescence has been addressed in multiple ways, including by rule-based transformation techniques. In such scenarios, a set of rules is executed against an artefact from a source language, generating a consistent or compatible artefact in a target language. For large and complex artefacts, and large and complex languages, this rule set may also be large and complex. Techniques that can make the execution or expression of these rule sets more efficient and effective would be helpful. In this paper we present a preliminary investigation on the use of preprocessing in software migration. In particular, we describe an approach that makes use of a novel cluster-based preprocessing algorithm, based on machine learning, to partition the input space, with the consequence of making rule-based transformation more efficient and effective. We present preliminary data evidencing this claim.