Advanced Metering Infrastructure (AMI) wireless communications network is an important part of smart grid architecture. How to effectively deploy fewer concentrators to collect data from smart meters in time will be a challenging problem. In the actual deployment, all we know are the meter and candidate concentrator position. We don’t know the link between meter and candidate is connectable or not before deployment or actual measurement. As the increases of number of candidate concentrator position, the measurement cost also raise. Therefore, this thesis will propose a feature extraction method from open data. With some test field data to training supervised classifier, we can predict the link is connectable or not before deployment.