{"id":258,"date":"2022-04-11T09:00:00","date_gmt":"2022-04-11T09:00:00","guid":{"rendered":"https:\/\/www.lancaster.ac.uk\/stor-i-student-sites\/danielle-notice\/?p=258"},"modified":"2024-03-10T16:44:42","modified_gmt":"2024-03-10T16:44:42","slug":"metric-learning-for-simulation-analytics","status":"publish","type":"post","link":"https:\/\/www.lancaster.ac.uk\/stor-i-student-sites\/danielle-notice\/2022\/04\/11\/metric-learning-for-simulation-analytics\/","title":{"rendered":"Metric Learning For Simulation Analytics"},"content":{"rendered":"Reading Time: <\/span> 5<\/span> minutes<\/span><\/span>\n

Usual output analysis of simulations, which is done at an aggregate level, gives limited insight on how a system and its performance change throughout the simulation. To gain greater insight regarding this, you can think of a simulation as a generator of dynamic sample paths. When we consider that we are in the age of “big data”, it’s now pretty reasonable to keep the full sample path data and to explore how to use it for deeper analysis. This can be done in a way that supports real-time predictions and reveals the factors that drive the dynamic performance. <\/p>\n\n\n\n

In this post, we’ll look at the emerging field of simulation analytics<\/strong>.<\/p>\n\n\n\n

  1. What is simulation analytics?<\/li>
  2. Metric learning for simulation<\/li>
  3. A simple example<\/li>
  4. Some final thoughts<\/li><\/ol>\n\n\n\n

    1. What is Simulation Analytics?<\/h2>\n\n\n\n

    The idea of simulation analytics<\/em> was first described by Barry Nelson<\/a>. It is not just “saving all the simulation data” and then applying modern data-analysis tools. It explores the differences between real and simulated data. Nelson outlines that the objectives of simulation analytics are to generate the following:<\/p>\n\n\n\n

    1. dynamic conditional statements<\/strong>: relationships of inputs and system state to outputs; and outputs to other (possibly time-lagged) outputs.<\/li>
    2. inverse conditional statements<\/strong>: relationships of outputs to inputs or the system state<\/li>
    3. dynamic distributional statements<\/strong>: full characterization of the observed output behaviour<\/li>
    4. statements on multiple time scales<\/strong>: both high-level aggregation and individual event times<\/li>
    5. comparative statements<\/strong>: how and why alternative system designs differ<\/li><\/ol>\n\n\n\n

      2. Metric Learning for Simulation<\/h2>\n\n\n\n

      The remainder of this post is a discussion of the work done<\/a> by one of my STOR-i colleagues, Graham Laidler and his supervisors.<\/p>\n\n\n\n

      We can use sample path data available to build a predictive model for dynamic system response. In particular they use k-nearest-neighbour<\/strong> classification <\/strong>of the system state with metric learning to define the measure of distance [1]<\/a> . In kNN classification, a simple rule is used to classify instances according to the labels of their k nearest neighbours. <\/p>\n\n\n\n

      From this definition, the paper uses <\/p>\n\n\n\n