Latest Contributions by JaySandhu

Location-Allocation does not "move" an existing facility. It chooses the best set from a larger set of facilities. So having a good candidate set is important. And you can choose to load in the existing fire stations as candidates (not required). If you set them as required then the same location will be chosen. But if you set them as candidate and solve then you may get some original locations chosen (meaning they are in optimal location). And you solve for the number of fire stations you want. If you had 10 existing and you still want 10 then solve for 10. This is a very subjective decision as slight change to the problem setup can give you a different set of "optimal" locations. So in case you end up suggestions to move, you still have to evaluate how much "better" service/response time you will get. So you should solve with the existing fire stations and then solve for moving them and compare. All LA helps you with is generate alternatives that you need to evaluate in the context of some larger goal. Not sure if I answered your question! Jay Sandhu ... View more

Since you have 300 demand points and 120 candidate facilities and want to choose 30, each with a capacity of 10, use the Maximize Capacitated Coverage problem type and set the Default capacity to 10 and number of facilities to choose as 30 and solve! Jay Sandhu ... View more

Why not simply multiply the demand weight by 2 to double it's importance to the facility. You can come up with a weighting factor based on how close it is to the water, etc., and use that weight factor to multiply the existing demand weights. Jay Sandhu ... View more

The GroupName field is used for Demand Points to make sure that the points with same name will be served by the same facility. But you cannot specify the exact facility that will be chosen to server them. The solve will figure out the best facility. You can however make your existing facility "required". That way the required facilities have to be in the solution and this MAY give you the correct answer if your existing facilities are near the optimal locations (as the demand groupname will force the entire group to be served by one facility). Jay Sandhu ... View more

Thanks for providing the background on what you are doing. In general retail stores have no capacity and thus the maximize market share does not try to limit that. Also, it is not working on "demand" to which one can apply the notion of capacity. What is is working on is the probability of a spatial interaction between the store and customer location based on distance away and the distance decay (power or exponential) that has been calibrated from some known behavioral analysis for the area and the product mix. So the only thing that will limit the interactions is the impedance cutoff. e.g., Customers do not come from more than 10 Km away. Sorry there is no notion of capacity in this case. As far as demand point being census group  points, all I was suggesting before is you can take the same location and add it again but halve the number of people that are there. So the total amount of people at the location is the same except they are now split into two separate points that can be allocated separately if you use the max capacitated coverage solver. ... View more

Perhaps you can describe what you are trying to achieve. Maximize market share has a different objective then the maximize capacitated coverage problem type. So if capacity is important you need to split the demand into smaller demand points and allocate that way. Jay ... View more

You mention a few things that are not clear: >Also looking at the time elements and time windows, the Out of the box Network Analyst doesn't seem to support the daily routing of a fleet, each day, for a 30 day window. I admit I am slightly perplexed as to how to implement the time windows when you can enter a starting time but no end time. In the VRP solver, you can enter as many routes/vehicles as needed. So lets say you have two vehicles and need to keep them busy for 30 days, then add 60 routes/vehicles as Vehicle1Day1, Vehicles2Day2, Vehicle1Day2, etc. You can set up the start/end times for the vehicles to limit it 8 hours. Now when you solve you have work for each vehicle for each of the 30 days. Each order can have two time windows. Each time window has a start and end time. See the Order Properties here: http://desktop.arcgis.com/en/desktop/latest/guide-books/extensions/network-analyst/vehicle-routing-problem.htm#GUID-14BF… As Joe mentioned, this may require a different solution (Chinese Postman) as opposed to the VRP solver in the Network Analyst extension. But I suggest that you try to solve this using the VRP solver and then evaluate the results. Model your problem as 5811 orders, one warehouse, one vehicle for each of the 30 days and solve. Jay ... View more

Gurmeet, can you post some more details on the size of your network dataset, and is your TEMP folder in the same place with the free 13 GB disk space. Jay Sandhu ... View more

If you just need that statistics on how the demand is allocated then you can open the attribute table for the Facilities and for each chosen facility it will tell how many demand points and amount of demand was allocated. If you want ONE demand point with say 10 demand weight allocated to more than one facility while using the Maximize Capacitated Coverage then you need to duplicate the demand point. So in this case you could put in 5 points with demand 2 or 10 points with demand 1 and then run the solver. Jay Sandhu ... View more

You are right that it is "all or nothing" allocation for each demand point. Besides the approach you mentioned about creating random points inside the census polygon, you could add the same demand point multiple times with less demand. Let's say you have a demand location with 100 units of demand. You could add that same location ten times with a a demand of 10. These points will be sitting on top of each other but they may be allocated to different facilities as needed. Jay Sandhu ... View more

Are you re-building the network dataset? It is possible that the locations of the points are no longer accurate because the source feature id's may have changed by the build's OR you had some older network location fields and the load locations used them instead of geometry as a way to locate the facilities. You can also "see" where a facility is located by selecting one in the NAWindow and press the number 1 key. It will flash the snapped location on the map. In the NAWindow you can right-click on the Facilities and do relocate all. And then resolve the service area again the see if the problem goes away. ... View more

You say you have 171,126 demand and 2624 facilities. The first thing that comes to mind is that it may not have sufficient RAM to complete this work. Do you specify any distance cutoff to limit the amount of demand each facility needs to evaluate? In case you have a 64 bit OS, can you use background 64 bit GP to solve the problem and see if you get an answer? Jay Sandhu ... View more