The Random Utility Maximization model is by far the most adopted framework to\nestimate consumer choice behavior. However, behavioral economics has provided\nstrong empirical evidence of irrational choice behavior, such as halo effects,\nthat are incompatible with this framework. Models belonging to the Random\nUtility Maximization family may therefore not accurately capture such\nirrational behavior. Hence, more general choice models, overcoming such\nlimitations, have been proposed. However, the flexibility of such models comes\nat the price of increased risk of overfitting. As such, estimating such models\nremains a challenge. In this work, we propose an estimation method for the\nrecently proposed Generalized Stochastic Preference choice model, which\nsubsumes the family of Random Utility Maximization models and is capable of\ncapturing halo effects. Specifically, we show how to use partially-ranked\npreferences to efficiently model rational and irrational customer types from\ntransaction data. Our estimation procedure is based on column generation, where\nrelevant customer types are efficiently extracted by expanding a tree-like data\nstructure containing the customer behaviors. Further, we propose a new\ndominance rule among customer types whose effect is to prioritize low orders of\ninteractions among products. An extensive set of experiments assesses the\npredictive accuracy of the proposed approach. Our results show that accounting\nfor irrational preferences can boost predictive accuracy by 12.5% on average,\nwhen tested on a real-world dataset from a large chain of grocery and drug\nstores.\n