As software cost drivers are fuzzy and uncertain, software cost estimates are prone to a certain degree of estimation errors especially in their early stages of software development life cycle. However, most of the existing software cost estimation models in present literature only generate a single point estimate and do not explicitly reveal the degree of risks caused by their inaccuracies. This paper proposes a fuzzy decision tree approach for embedding risk assessment information into a software cost estimation model. Using this model, one may be able to determine the software cost estimate as well as the estimation error in the form of a fuzzy set. In verifying the merits of this model, we have used the 63 historical project data in the COCOMO model. The validation result shows that our proposed model reveals the risk assessment of the generated software cost estimate, and at the same time yields an even more accurate result as compared to the original COCOMO model.