Stock price prediction, a classic challenge in the realms of finance and computer science, has attracted a multitude of scholars and investors over the years, leading to the development of diverse prediction methods, theories, investment strategies, and practical experiences. Analyzing text data from social media, alongside numerical indicators, can assist in forecasting future stock price movements. In our study, we integrate fuzzy set theory into LS-KSVCR to create FH-LS-KSVC, a novel fuzzy support vector machine utilizing fuzzy hyperplane for K-class stock market trend prediction using social media data. FH-LS-KSVC assigns membership degrees to data samples based on training sample importance and employs fuzzy numbers for optimal hyperplane components. This fuzzy hyperplane effectively captures imprecise real-world characteristics, reducing the impact of noise. Experimental results in real stock trend classification applications demonstrate that FH-LS-KSVC combines the benefits of LS-KSVCR for multi-category classification performance and FH-SVM for noise robustness.