Abstarct: Detection of gas chimneys provides an impressive approach for the direct hydrocarbon identification, due to the path of gas migration from the reservoirs can be revealed by gas chimneys. Seismic attributes are the best tool for gas chimneys detection in seismic data. Seismic attributes are the useful tool that help the interpreters in the analysis of reflection seismic data for better identification of gas chimneys, faults, and buried channels, reefs, thin laxyers, salt domes, buried deltas, etc. Each seismic attribute cannot lonely able to accurately identify the gas chimney. Therefore, the combination of several attributes is used to achieve the gas chimney detection. We used the multi-laxyer perceptron neural network (MLP) to classified the seismic image baxsed on the multi attributes fusion. On the other hand, using a large number of attributes is time-consuming and confusing, and for this reason, the feature detection process is needed to run before the data integration. We proposed an optimization algorithm named as biogeographical baxsed optimization (BBO) for feature selection. The BBO optimization algorithm selects optimal attributes using its two main approaches, i.e., migration and mutation. The proposed algorithm was applied to the F3 block data set and the BBO algorithm selected 15 attributes from the initial 79 attributes. Finally, MLP has been used to detect and separate gas chimney from non-gas chimney with an error of less than 1.3%.