Study of linear regression based on least squares and fuzzy least absolutes deviations and its application in geography

In regression models normally, both of data and parameters are considered as crisp. But, in some cases, for improving the prediction, we need to prepare and use a regression model with imprecise coefficients. In this case the normal regression models are not suitable, so fuzzy regression can be fair replacement models. In this paper we consider the least square and least absolute deviation familiar methods to compare the mention models. Finally we apply these approaches to geography data (TMP, PRC, Latitude and Longitude) with symmetric fuzzy observations.