This paper employs linear regression algorithms in order to train models under the presence of limited training data. Usually in transportation applications, these models are built via Ordinary Least Squares and Stepwise Regression, which perform poorly under limited data. The algorithms presented in this paper have been extensively used in other scientific fields for problems with similar conditions and seem to partially or fully remedy this problem and its consequences. Four different algorithms are presented and several models are built. The models are used for truck volume prediction on highway sections in New Jersey, and results are compared to Stepwise Linear regression models.
