Many applications require a careful selection of attributes or features from a much larger set of data. This attributes selection problem need to optimized. In order to tackle this problem this paper proposes a binary-real code multi-gen multi-parameter genetic algorithm for attributes selection from large seismic data and prediction of effective porosity. Genetic Algorithm (GA) uses three selection methods for this purpose, mean square error and correlation coefficient are two witness criteria to choose the best subset of attributes that minimize the error and give high prediction of porosity.