SQL Server Data Mining: Plug-In Algorithms

Microsoft SQL Server Analysis Services 2000 Service Pack 1 allows the plugging in (“aggregation”) of third-party OLE DB for Data Mining providers on AnalysisServer. Because this aggregation is at the OLE DB level, third-party algorithm developers using SQL Server 2000 SP1 have to implement all the data handling,parsing, metadata management, session, and rowset production code on top of the core data mining algorithm implementation.By contrast, SQL Server 2005 Data Mining allows aggregation directly at the algorithm level. Although this restricts what the third-party algorithm developercan support in terms of language and data types, it frees the developer from implementing all the additional layers described above. It also allows for muchdeeper integration with Analysis Services, including the ability to build OLAP mining models and data mining dimensions. We use the term “plug-in algorithms” todescribe third-party algorithms that plug into the SQL Server 2005 Analysis Server (hereafter referred to as “Analysis Server”) and appear, in all respects, likenative algorithms to users.Describes how SQL Server 2005 Data Mining allows aggregation directly at the algorithm level. Although this restricts what the third-party algorithmdeveloper can support in terms of language and data types, it frees the developer from having to implement data handling, parsing, metadata management,session, and rowset production code on top of the core data mining algorithm implementation.