<![CDATA[Mental health issues among college students are a critical issue that requires data-driven approaches to detect early treatment needs. This study aims to analyze and compare the performance of three machine learning algorithms: Naive Bayes, K-Nearest Neighbor (K-NN), and Decision Tree in classifying college students' mental health treatment needs based on an open survey dataset. The study was conducted systematically using RapidMiner software, with data preprocessing, model training, testing, and performance evaluation using accuracy, precision, and recall metrics. The test results showed that the Naive Bayes algorithm produced an accuracy of 78.85%, a precision of 75.96%, and a recall of 72.84%. K-NN performed better with an accuracy of 82.62%, a precision of 80.83%, and a recall of 77.37%. Meanwhile, the Decision Tree algorithm performed best with an accuracy of 88.32%, a precision of 86.77%, and a recall of 85.80%. In addition to its high performance, Decision Tree also offers advantages in interpreting results through its decision tree structure, which illustrates the role of variables such as employment status (self_employed), family history (family_history), survey completion time (timestamp), and care options (care_options) in the classification process. Decision Tree can be concluded as the most effective classification model for detecting student mental health needs in this data context. These findings are expected to serve as a reference in the development of machine learning-based early detection systems to support mental health policies and interventions in higher education settings.]]>
