Algorithm 1. Construction of initial parameters

Input: training dataset, clustering parameters $K,\delta ,\varsigma$, maximum iteration, density parameter $\gamma$, result reference values Output: reference intervals $I_j$, representative points within reference intervals $I_k^R$, initial belief degrees ${\beta }_{l,k}$ Procedure: 1: Load training dataset 2: Extract dataset, antecedent attributes values and result values 3: ▷ FCM-based reference interval partition 4: for each feature f = 1, …, F do 5.  for K = $K_{\min },\dots K_{\max }$ do 6:     Run FCM on dataset and compute BIC 7:  end for 8:  Select optimal $K$ with minimum BIC 9:  Generate reference intervals 10: end for 11: ▷ DWDS key point selection 12: for each interval ${\rho }_i$ of feature do 13:   Compute representative density ρ and distance to center 14:   Select point with maximum score $S_i={\displaystyle \frac{{\rho }_i}{1+V_{\mathrm{i}}}}$ 15: end for 16: ▷ Belief degree calculation using GIBM 17: for each training sample do 18:   Compute output membership based on result reference values 19:   Update rule belief degrees using Gaussian activation 20: end for 21: Normalize belief matrix ${\beta }_{l,k}$ 22: Extract reference intervals $\to I_j$ 23: Extract representative points $\to I_k^R$ 24: return $I_j$, $I_k^R$, ${\beta }_{l,k}$