The complexity and variety of bot behaviors on social media platforms like X (formerly Twitter) demand advanced detection methods that can handle multiclass imbalances effectively. Existing binary classification methods often fall short in accurately identifying and distinguishing between various bot types and genuine users, leading to biased and incomplete detection. To address these challenges, we introduce HyperSMOTE-MC, a novel hypergraph-based approach specifically designed for multiclass bot detection. By constructing a hypergraph where users are represented as nodes and their interactions as hyperedges, HyperSMOTE-MC captures the multifaceted relationships among users. This method employs synthetic minority oversampling to balance the dataset, ensuring fair representation of all bot classes. Additionally, HyperSMOTE-MC integrates a Hypergraph Convolutional Network (HGCN) to leverage these complex interactions for improved classification performance. Evaluated on the TwiBot-20 dataset, HyperSMOTE-MC demonstrates superior accuracy, precision, recall, F1 score, and AUC-ROC compared to baseline methods, showcasing its robustness and effectiveness in handling multiclass bot detection across various domains.