 Remember in the middle school dance example, we started out with just randomly placing the location of our clusters, the centroids, and then we move them around trying to find a better location. The idea is that as we move the locations around and we start calculating the Euclidean or crowfly distances from the individuals to the centroids, we want to minimize those differences, those distances, until we come up with the optimum location for the centroids which minimizes the total distance from all of the individuals in that cluster to that centroid. And that's what we're going to do now using Solver. We've got our dance hall here, our space, these are our clusters, and for each dimension we will have a location for that cluster and also the location of the customers being either zero or one if they are participating in that particular space, that particular dimension. And because we're using zero and ones, we want to control the locations for the clusters to be greater than zero and less than one. And if you remember, we use Solver in an earlier problem to help us come up with an optimum solution of either maximizing an objective function or minimizing one or hitting some value. Here we're going to use this total distance which is the sum of the minimum distance to a cluster for each of the customers and use that as our objective function. Go to the data tab, click on Solver, open up the Solver dialog box. The first thing we want to do is set our objective, we click on that empty cell there to make sure it's blinking and then go down and click on our total distance and then we want to minimize it and we're going to change the locations of the clusters and we click in that box and start with cluster one and we go all the way down to K33, which gets all of the clusters for all of the offers. Now we need to add constraints. I mentioned we need to keep the variables here between zero and one. So let's add a constraint and our cell references again from one to K33 and we want them to be less than one and I'm going to click OK. If you notice in the Solver if we've got make unconstrained variables non-negative that means that these variables cannot go below zero. Now we need to change the solving method to something called evolutionary. We need to set the options for evolutionary so we open up Solver's options box and find it there, click on evolutionary tab, make sure we've got required bounds on variables. We want the convergence to be .0001 for zeros, mutation rate we want to be 0.015, population size 100, random seeded default zero and change the maximum time that improvement to 600 and again require bounds. So we click OK and then we click solve. Now this is going to take a while and I'm not going to wait the whole time. Depending upon how fast your computer is, how many programs you have running, how much RAM you have, it could take 20 or 30 minutes. If you think about it we have an awful lot of things that Solver has to check in order to try to minimize this distance. So it's going to take a while. It goes through and tries solution, moves around these values and the locations and then solves all those equations simultaneously trying to figure out how to minimize this total distance. So I'm going to pause the video here and come back once we get a solution. I might mention if you look down at the very bottom of your worksheet you can see that Solver is working there. Right now it's found an incumbent. It calls it 162.79 and it's on subproblem 1200 and I don't know how many different solutions has tried. So it's cranking away. Look down there to make sure it is moving just to give you some reassurance that you're not wasting your time on the solution. So again I want to pause. Okay it took about 30 minutes of my machine. I had some other programs running I've forgotten about but it got a minimum total distance of 140.54 and of course whatever you're doing would probably have a different objective but that would give you an idea. Here are the cluster locations and that's kind of cluttered. I want to change the formatting there to make it a little bit more understandable home and let's reduce the number of decimal places down so we can get something a little more intelligent. Whoops, meeting before. Okay and now if we scroll down to the bottom we'll see that we have an assigned cluster for each one of our customers. Two for Adams, four for Allen, one for Anderson, two for Bailey. Eric begs the question why did Solver assign them to those clusters? And we'll take a look at that next.