Optimal Call – Record Sampling for Network Planning and Administration

We use optimal systematic sampling techniques to improve dramatically the call-record sampling process for network planning applications. Although we take a fairly traditional approach to sample optimization, our objective function is unusual, our solution algorithm is original and efficient, and the application to network planning problems is unique. Moreover, we show that, relative to results for existing techniques, we can reduce average root mean squared errors of customer calling volumes by a factor of 1.5—9 for randomly generated problems and by a factor of 2.5 for a major metropolitan area. These error reductions are expected to translate into significant savings in network costs, improvements in service quality, or increases in network revenues.