Data Mining and Visualisation Infrastructure • Aimed to address the overlap between visualisation and data mining and the data, and the infrastructural requirements. – found this difficult after we lost our data-mining expert (Chris) … particularly given our first question (just after he left) was: • What is the difference between data mining and visualisation? NESC Scientific Data Mining, Integration and Visualisation Breakout Group Feedback (25 Oct 02) What is the difference between data mining and visualisation? • data mining is machine oriented, visualization is human oriented • data mining is number crunching on assumption that some model can be used; visualization is more exploratory data analysis. Also we use visualization to validate the model • visualization is presentation with added value? – vis is useful for generating hypothesis; and for testing hypothesis. • Considerable overlap in functionality (from the data-flow point of view), but what about overlap in implementation, interfaces? NESC Scientific Data Mining, Integration and Visualisation Breakout Group Feedback (25 Oct 02) What standards exist? • Why? To allow communication between processes. – (How to plug data into data-mining -> visualisation chains?) • Answer? None! What are people currently doing? • For data discovery: DAME and NDG are both using Z39.50 now and planning to use OGSA/DAI … – Z39.50 provides a protocol for indexing databases for distributed queries. – still not obvious what OGSA/DAI will give us beyond Z39.50 and that’s here now … – … but Z39.50 doesn’t do more than using HTML hyperlinks to hand over from discovery to data (it’s mostly used for indexing discovery information not the information/data itself!) – The Open-Archive project has an alternative approach to either, which involves harvesting metadata, but might share the same hand-off problem? (In any case need to talk more with library community). NESC Scientific Data Mining, Integration and Visualisation Breakout Group Feedback (25 Oct 02) What next? The challenges! • Standards, standards, standards! • At the moment those who are not in the “in-crowd” can ask the following question “how do I use these tools (which tools?) with my data? • Can we describe what a particular data mining tool can do? Can we describe what inputs it requires? Can we describe it’s outputs? Can any of these descriptions become machine readable? • The challenge is to address these points!!! NESC Scientific Data Mining, Integration and Visualisation Breakout Group Feedback (25 Oct 02) NESC Scientific Data Mining, Integration and Visualisation Breakout Group Feedback (25 Oct 02)