The implementation of latest International Maritime Organization emission standard raised stringent requirements for marine domestic sewage discharge. In this study, an air-lift multilevel circulation membrane reactor (AMCMBR) was operated to analyze effects of various ecological factors on effluent of marine domestic sewage. Back-propagation (BP)-Artificial Neural Network (ANN) was used to simulate effect of each ecological factor on reactor performance. The activities of four enzymes were investigated to reveal microbial activities in reactor. Experimental results indicates that the Hydraulic Retention Time (HRT), Mixed Liquid Suspended Solids (MLSS) and pH value cannot be less than 4 h, 3000 mg/L and 6, respectively to meet the IMO emission standard for effluent COD. A small value of mean square error (0.00147) indicated that BP-ANN can well describe the relationship between operation parameters (influent COD, HRT, MLSS, and pH) and effluent COD. The order of relative importance was pH ≈ MLSS > HRT > influent COD. Polyphenol oxidase and urease can serve as indicating factors for reactor performance, whereas dehydrogenase and nitrate reductase showed less susceptible towards varied influent COD and MLSS.