Text Classification of Facebook Messages Using Multiclass Support Vector Machine

Text classification is assigning classes or categories to a specific set of words. Different text classification techniques are used in classifying academic-related messages with acceptable accuracy. However, using datasets about academic-related messages from previous studies, which are formal in structure, will provide inaccurate results when used for training since the messages from Facebook pages are informal. This study focused on creating a classifier that can categorize informal academic-related messages from the Facebook pages of schools of Mapua University. The dataset collected undergoes Data Pre-processing to clean the messages. Then, the cleaned messages are converted to vectors using the TF-IDF Text vectorizer. Four SVM kernels were tested for text classification: Linear, Sigmoid, RBF, and Polynomial. Sigmoid achieved the highest accuracy of 81.6% and F1-score of 81.4 in classifying informal academic-related messages.