A number of methods have been proposed to build a relationship between effort and size. These models are generally based on regression analysis and a widely accepted model is not yet available. Although in some sizing methods, such as MKII and IFPUG, different multipliers for the base functional components (BFC) exist, their origin and the purpose of their usage are undefined. The COSMIC method does not treat components separately and assigns the same measurement unit to each of them. In this study we used the Artificial Neural Network and regression based methods to create effort estimation models that take the four components of the COSMIC method into consideration. In the research we compared several functional size based effort models in terms of accuracy using a reliable company dataset. These models comprised not only the generic models proposed in the literature or currently in use, but also specific models that we generated using our dataset with a single and multi-variate regression analysis and the ANN method. We also explored the effect of functional similarity (FS) using our specific models. We found that using BFC instead of total size improved effort estimation models and the ANN method is a useful approach to calibrate these components according to the company characteristics.