Students' mental health is an important issue that needs serious attention, especially in the era of social media which is full of psychological pressure. This research aims to develop EduMood, a web-based sentiment analysis system to monitor college students' mental health issues by analyzing tweets on Twitter. The tweet data is collected using relevant keywords and goes through preprocessing stages such as text cleaning, bilingual lexicon-based initial labeling, and balancing the amount of data between sentiment classes. The system uses two machine learning algorithms, Support Vector Machine (SVM) and Naive Bayes with Term Frequency-Inverse Document Frequency (TF-IDF) feature representation. The evaluation results show that SVM has a higher accuracy of 99.3% compared to Naive Bayes which reaches 96.5% with f1 scores for all classes above 0.99 for SVM. EduMood is implemented as a web-based application using Flask and Bootstrap 5, which presents the analysis results through an interactive dashboard. The dashboard displays the aggregate sentiment distribution in the form of diagrams, wordclouds, monitoring tables, and text manual predictions. The results of this study show that EduMood not only provides excellent model performance, but also offers a practical solution for the campus to monitor the psychological condition of students in a fast, real data-based, and easily accessible manner. This system is expected to support efforts to improve student mental health in a sustainable manner.