Geomagnetic super-storms of October and November 2003 are compared in order to identify solar and interplanetary variables that influence the magnitude of geomagnetic storms. Although these superstorms (DST < -300 nT) are associated with high speed CMEs, their DST indices show large variation. The most intense storm of November 20, 2003 (DSt∼ - 472 nT) had its source in a comparatively small active region and was associated with a relatively weaker, M-class flare, while the others had their origins in large active regions and were associated with strong X-class flares. An attempt has been made to implement a logistic regression model for the prediction of the occurrence of intense/superintense geomagnetic storms. The model parameters (regression coefficients) were estimated from a training data-set extracted from a data-set of 64 geo-effective CMEs observed during 1996–2002. The results indicate that logistic regression models can be effectively used for predicting the occurrence of major geomagnetic storms from a set of solar and interplanetary factors. The model validation shows that 100% of the intense storms (-200 nT < DSt < -100 nT) and only 50% of the super-intense (DST < -200 nT) storms could be correctly predicted.