Linear Regression is used as a prediction tool in transportation planning, traffic data analysis and safety. Many researchers have attempted to address several transportation issues using these types of models. The accuracy and stability of these models is mainly dependent on the size of the available data. Unlike other science fields, where sophisticated algorithms are used to deal with problems of small datasets, the majority of freight demand modeling relies on simple statistical techniques. Limitations of these modeling methodologies, caused by their high dependence on data availability, and several assumptions that need to be made can result in erroneous models. In cases in which limited data are available, more advanced algorithms that can be legitimately used on small datasets should be applied. In this paper a description and classification of algorithms and processes used for creating these types of models under limited data is presented. To demonstrate the applicability of these algorithms, along with implementation problems, limitations, and the different performance measures, a case study is used. Different models are created and results are presented and discussed.