In this letter, a non-reciprocal filter with enhanced directivity is analyzed methodically and the filter parameters are optimized using an evolutionary algorithm. The return loss, insertion loss, and isolation characteristics of the filter exhibit a trade-off that makes manual tuning a trial-and-error method. The veracity of the numerical modeling is conformed by designing a 150 MHz lumped element non-reciprocal bandpass filter based on the parameters extracted using an evolutionary algorithm based particle swarm optimization (PSO). The simulated and measured results comply well with the modeling and the results exhibit maximum directivity of 28.2 dB without degradation in insertion loss (1.1 dB) and return loss (16.2 dB) within the passband. The algorithm can be utilized in designing non-reciprocal filters having different center frequencies and bandwidths.