Chondrosarcoma of the Pelvis is one of the most challenging tumors for the orthopaedic surgeon. Clinically, imaging modalities represented by Magnetic resonance imaging (MRI) are the regular approach for diagnosing Chondrosarcoma. In this paper, to further improve the automation of diagnosis and mine the complementary clinical information from high-dimensional features, we propose a pelvic chondrosarcoma identification model based on MRI radiomics features. Specifically, radiomic features, such as first-order, shape, and texture, are initially extracted, and then a subset of these features are selected based on the least absolute shrinkage sum selection operator (LASSO) analysis. Subsequently, a recursive feature elimination (RFE) algorithm is utilized to select the most significant features. The prediction results are finally obtained through the constructed multivariate logistic regression model. Experiments conducted on clinical data achieved the area under curve (AUC) of 98.1% on testing set, demonstrating the effectiveness of the proposed MRI-based model for identifying pelvic chondrosarcoma.