In this paper, we propose a new convex combination weight rule for the cross decomposition method which is known to be one of the most reliable and promising strategies for the large scale optimization problems. It is called generalized cross decomposition, a modification of linear mean value cross decomposition for specially structured linear programming problems. This scheme puts more weights on the recent subproblem solutions other than the average. With this strategy, we are having more room for selecting convex combination weights depending on the problem structure and the convergence behavior, and then, we may choose a rule for either faster convergence for getting quick bounds or more accurate solution. Also, we can improve the slow end-tail behavior by using some combined rules. Also, we provide some computational test results that show the superiority of this strategy to the mean value cross decomposition in computational time and the quality of bounds.