Predicting the sensitivity class of loess soils using ordinal logistic regression method, Northeast of Iran

The current study evaluates ordinal logistic regression (OLR) for assessing collapse sensitivity classes of loess soils. Collapse sensitivity (Is) is the critical parameter to predict pseudokarst sinkholes occurrence in Golestan Province in northeastern Iran. A database containing 62 records of soil's physical and mechanical properties is used in this study. By performing oedometer tests, the parameters of collapse coefficient, the time required for 90% settlement (T90%), and collapse sensitivity were determined. To gather this goal, a database was prepared based on experimental datasets, consisting of ten inputs (grain size analysis, porosity, initial water content, precipitation, climate, liquid limit, calcium carbonate, vegetation, and degree of soil saturation) and one output (collapse sensitivity classes). This task is complex due to the difficulty of preparing and carrying out such experiments in a laboratory. Using OLR, the probability of soils being placed in classes with severe, moderately severe, moderately, and slight sensitivity was estimated. This study showed that the OLR method could correctly distinguish more than 70% of different categories. Experimental data obtained from Semnan, Sarakhs, and Mashhad areas, has shown the accuracy of the proposed OLR model.