Emissions are one of the key performance metrics for combustion systems. The effect of a forced dilution air jet, introduced through the combustor shell, on the nitric oxide emissions in the combustion chamber has been investigated experimentally. Afterwards, the combustor behavior in terms of NO emissions has been modeled by utilizing artificial neural networks. The parameter space is wide enough to train the network. Equivalence ratio varies between 0.7 and 1.0, blowing ratio between 0 and 16, and forcing frequency between 0 Hz to 850 Hz. A simple multi-layer-perceptron network with two hidden layers captures the plant behavior with sufficient accuracy and has a good generalization capability. Once trained on an actual combustor, this network can then be utilized to optimize combustor behavior, when coupled with an extremum-seeking control algorithm. Consequently, the side air jets can be modulated with optimal frequency that minimizes emissions. This approach has a potential implementation in industrial gas turbine engines.