Abstarct: In mechanized longwall mining proper caving of immediate roof guarantees the success of operation while delayed or poor caving will lead to severe consequences which results in reduction of safety and productivity. Therefore, a reliable evaluation of strata caving potential and prediction of the main caving span is imperative in the planning stage of longwall projects. The aim of this thesis is to present a qualitative and quantitative model to evaluate immediate roof strata cavability and prediction of the main caving span. For this purpose, nine inherent parameters were chosen as significant affecting factors on strata cavability. A fuzzy hybrid multi criteria decision making (MCDM) was used by combining fuzzy analytic network process (ANP) technique and fuzzy decision making trial and evaluation laboratory method (DEMATEL) to develop a new classification system. Subsequently, Roof Strata Cavability index (RSCi) was defined which qualitatively categorizes cavability level into five classes from uncavable to highly cavable. The qualitative model was developed in two forms baxsed on two different approaches including high accuracy and easiness of the application. Validation of RSCi with the use of twelve actual cases of worldwide longwall panel showed logarithmic and exponential functions delineate relationship between the main caving span and RSCi of first and second approach, respectively. In addition, capabilities of RSCi was investigated to predict and model the main caving span. The results showed that the linear model baxsed on RSCi of the first approach and hydraulic radius (HR) width possesses a higher performance in the prediction of the main caving span in actual cases when compared to the other models. In order to develop quantitative model, discrete numerical simulation incorporating distinct element code UDEC software was applied. To that end, at first, parametric study was conducted to investigate the role of seven significant parameters in the main caving span value. The results showed that the main caving span has direct relationship with equivalent immediate roof strength (EIRS) and GSI, has inverse relationship with lamination spacing, mining depth and extraction height and does not has a clear relationship with in situ stress ratio and advance rate. In order to establish an initial databaxse, the main caving span was modeled for different values of RSCi, extraction height and GSI. Accordingly, initial databaxse was extended to 738 datasets with considering 15% overestimate and underestimate in numerical modelling as well as a COA-SVR method. The value of the main caving span for each class of RSCI was analyzed probabilistically and probability density function (PDF), cumulative distribution function (CDF) and statistical indices was found out. Subsequently, the comprehensive model was proposed as a table in which the range of the main caving span for each RSCi class was proposed in 75% confidence interval. The results of model validation show that 75% of actual cases have the main caving span in the proposed range. It was concluded that proposed models are simple and efficient tools to evaluate the cavability of immediate roof and to predict the main caving span in mechanized longwall mining.