Fuzzy regression based on least absolute deviations is investigated. First, we introduce a metric,, on the space of LR fuzzy numbers which measures the distance betwee such fuzzy numbers based o orm. In order to find the coefficients of fuzzy linear regression models, we minimize the sum of distances between observed and estimated fuzzy outputs for all observations. For simplifying the minimization problem, we translate it to a standard mathematical programming problem. After introducing some indices of goodness of fit of fuzzy regression models, we assess the performance of the proposed models by several common data set.