Abstract: Abstract: Nowadays, a large number of land reclamation projects have been carried out or in the process of construction in coal mining areas of Xinjiang, and it is necessary to build up classification index and discriminant model to evaluate the suitability of land reclamation, for reducing the destruction of land and sustaining the development of coal mine. This article focused on the research of the theory of suitability evaluation of land reclamation and established a discriminant model that was used for the evaluation of coal mining areas in Xinjiang. The main type of destructed land in Xinjiang was grassland and bare land, so the land reclamation suitability evaluation was aimed at grassland. The results of land reclamation evaluation were divided into the most suitable, moderately suitable, barely suitable and unsuitable. One of the core issues of land reclamation is to construct evaluation index system. Taking 15 coal mines in Xinjiang as the research subjects, based on Bayes discrimination analysis, the factors of suitability evaluation of land reclamation of mines were analyzed; combined with the characteristics of land destruction and ecological geography of coal mines in Xinjiang, 7 factors such as the slope of land reclamation unit after land destruction, the depth of land destruction or the deposit amount of land occupancy, the guaranteed rate of soil for land reclamation, groundwater depththe clay content in soil, the thickness of topsoil, river runoff for irrigation of land reclamation unit and the content of organic matter in soil were selected as the evaluation factors. Each evaluation factor quantization value was taken from the measured value and the related report data. The 76 land reclamation units of 15 coal mines were used as the study samples. The Q clustering analysis method was used to classify 76 samples. The clustering analysis used the method of average connection to define the distance of categories for merging samples, and the distance was measured by the squared Euclidean distance. The results showed that the best plan was the 4 categories. Compared with the code, the classification results of this method were more realistic and accurate. By taking the 76 land reclamation units of 15 coal mines in Xinjiang as the training samples, the Bayes discriminant model for suitability evaluation of land reclamation of coal mining areas in Xinjiang was established. This model was composed of 4 discriminant functions. Every sample was tested by using the resubstitution method, and the correct rate was equal to 98.7%. The discriminant model was applied to 2 coal mines in Xinjiang. The results showed that the classification results were entirely consistent with the neural network method and the combination of factor and clustering analysis. The classification results complied with the classification result range of the code and were more accurate and realistic. The results of the modelling test and calculation have proved that the structure of Bayes discriminant is stable, and the discriminant effect is ideal. This model has strong analysis ability for suitability evaluation of land reclamation in coal mining areas of Xinjiang. The biggest difference between the results of evaluation model and the code was that results of evaluation model are more specific and detailed than that of code, especially in barely suitable category and unsuitable category. The classification results of this model are more accurate. This evaluation model can be widely used in suitability evaluation for land reclamation of coal mining areas in Xinjiang.