Friday, May 1, 2009
More fun with learning
Yesterday and today i did some reading on processes used in assessing, attacking, and solving problems with Data Mining. One thing of particular interest that I just read is an overview of the way that Neural Networks work. I had already started to wonder if Data Mining could be implemented in Chemical Plants as a part of process control, and I just read that one of the best tools for creating predictive models of chemical plants is through a Neural Network. This makes sense, as literally thousands of variables, some which are outside of the plant operator's control, must be taken into account when optimizing a chemical process. Even though a neural network perhaps cannot understand the scientific reasoning behind why the correlations it is able to draw are valid, this does not seem to matter so far as the overall model is accurate. I am finding the idea of data mining as it applies to predictive modeling (I'm guessing that that is also related to machine learning) to be very interesting.
Subscribe to:
Post Comments (Atom)
Indeed, neural networks have been successfully used for plant monitoring. They have the advantage of being data-driven and thus sensitive to incremental changes to "normal" operating conditions. People who have worked in this area have also called their work novelty detection, where the idea is to discover behavior outside the norm (in the case of a chemical plant, such abnormalities may require human intervention to prevent damage and/or accidents).
ReplyDelete