A General Approach to Construct Intuitionistic Fuzzy Regression Model Based on a New Least Absolute Discrepancy

The data sufferance from uncertainty and vague events impose statistical regression modeling to be treated in fuzzy environment for more accuracy upon causal relationship between independent and dependent variables. For more informative fuzzy, among limited studies on intuitionistic fuzzy regression models (IFRMs), this study proposed the most general approach to construct a full IFRM (the model’s parameters and output as well as input variable(s) are represented as symmetric and/or asymmetric positive and/or negative and/or mixed of neither negative nor positive triangular intuitionistic fuzzy numbers (TIFNs)). The estimated parameters of the proposed IFRM determined on solving linear programming problems based on intuitionistic fuzzy least absolute of discrepancies (IFLAD). The proposed approach respects homogeneity principle in modeling such that the constructed IFRM for fitting symmetric TIFNs is symmetric either (the intercept and slope(s) are symmetric in this case). Furthermore, two illustrative examples are used to show the soundness and robustness of the proposed model and compared with existing IFRM.