Autoimmune diseases pose an increasing challenge to healthcare systems in Mexico due to their complexity and rising prevalence. In this context, machine learning and deep learning techniques have emerged as valuable tools to support early detection, diagnosis, and personalized treatment. However, their application faces a critical barrier: the scarcity of clinical data in Spanish, particularly the lack of open and structured datasets that reflect the Mexican context. To address this issue, this study proposes and evaluates five machine learning models for the automatic classification of medical text related to autoimmune diseases. We trained the models using different preprocessing strategies and semantic representations (FastText and GloVe) to analyze their impact on classification performance. Experimental results show that Model 1 achieved the best overall performance with an F \(_1\) of 0.77, followed by Model 5 (F \(_1\) = 0.76) and Model 4 using GloVe (F \(_1\) = 0.75). These findings suggest that combining traditional machine learning algorithms with enriched semantic representations significantly enhances classification accuracy, even when limited data are available.