The distribution utility is aware of the users’ connected load; nevertheless, their usage patterns remain unknown without the installation of smart meters on-site. Moreover, the components of the feeders at the primary distribution level remain unidentified. This research tackles this important problem by doing load profiling at the primary distribution level using data mining techniques. The method of least square error is used to model load profiles. The profiles were then clustered using the K-means algorithm and silhouette analysis into three main sectors. An optimization task is executed to divide the feeder load into percentages of clustered sectors.