In this work quantitative structure-activity relationship (QSAR) study has been done on 1,2-ethylenediamine derivatives as anti-tuberculosis drugs. Genetic algorithm (GA), artificial neural network (ANN), multiple linear regressions (stepwise-MLR) and Imperialist Competitive Algorithm (ICA), were used to create the nonlinear and linear QSAR models. The root-mean square errors of the training set and the test set for GA–ANN models using the jack-knife method, were 0.1402, 0.1304 and Q2 = 0.94. Also, the R and R2 values 0.85, 0.73 in the gas phase were obtained from a GA-stepwise-MLR model. Q2 of training set for PLS was 0.52. The results obtained from this work indicate that ANN and ICA models are more effective than other statistical methods and exhibit reasonable prediction capabilities. The best descriptors are G3u, HATS2e, F02(C-N), GGI10, RDF040m, Mor22p, Mor05p, TIC4, H4e, H-052, G2m and G1e.