This paper designs a siting scheme for public electric vehicle chargers based on a genetic algorithm working on charger monitoring streams. The monitoring-combined allocation scheme runs on a long-term basis, iterating the process of collecting data, analyzing demand, and selecting candidates. The analysis of spatio-temporal archives, acquired from the fast chargers currently in operation, focuses on the per-charger hot hour and proximity effect to justify demand balancing in geographic cluster level. It leads to the definition of a fitness function representing the standard deviation of per-charger load and cluster-by-cluster distribution. In a chromosome, each binary integer is associated with a candidate and its static fields include the index to the cluster to which it is belonging. The performance result obtained from a prototype implementation reveals that the proposed scheme can stably distribute the charging load with an addition of a new charger, achieving the reduction of standard deviation from 8.7 % to 4.7 % in the real-world scenario.