We propose an hybrid discrete choice framework for modeling decisions of investors performed on stock markets. Complex and detailed behaviors have been pointed out and quantified. We focus on the choice of action (buy or sell) and the duration until the next action. The choice of action is handled with a binary logit model with latent classes characterizing the perception of the risk, while a Weibull regression model is used for the duration until the next action. The duration model also accounts for the risk perception. Both models consider the dynamic nature of the underlying phenomenon. They are merged in a single model called combined model. It is estimated using data from a Swiss bank consisting of 25989 observations of transactions performed between January 2005 and September 2010, in 6 different funds. The predictive performance of the models are tested. A cross-validation analysis is performed. The forecasting accuracy of the action model is studied more in details. Parameters of both models are interpretable and emphasize interesting behavioral mechanisms related to investors' decisions. The good predictive capabilities of the action model in a real context makes it operational, which is a significant contribution.