This paper proposes, Data Envelopment Analysis (DEA) models and algorithms to reduce inputs and increase outputs for an inefficient Decision Making Unit (DMU) in a network DEA. Inputs and outputs considered for change are under constant sum constraint. In a network, DMU composed of multiple divisions, it is possible that changes in inputs and outputs of one division may affect the efficiency of other divisions. At the same time, it is also possible to change the efficiency of a division without changing the efficiency of the entire network. Models to reallocate inputs/outputs under constant sum constraint need to take these possibilities into account. The reallocation of inputs or outputs is done without reducing the network DEA efficiency of the other DMUs. Models and algorithms have been developed to transfer intermediate products from one DMU to another under constant sum constraint. These models and algorithms will help the decision maker to allocate resources more effectively among various divisions in order to improve the efficiency of the system consisting of these interrelated divisions. Theoretical results have been illustrated with the help of a numerical example. © 2015 IEEE.