A Machine Learning-Based Topic Extraction and Categorization of State Universities and Colleges (SUC) Customer Feedbacks

Academic institutions collect an increasing amount of data through surveys. Aside from the usual numeric ratings obtained from the survey, the hierarchical concerns and sentiments can also be identified through the text-based customer feedbacks. These feedbacks contain text about customer experiences with the products offered and services delivered by an institution. A challenge in analyzing unstructured customer feedback is in making sense of the topics that are expressed in words used to describe these experiences. This study develops a model for text analysis of the customer feedbacks that exploits machine learning algorithms such as topic modeling. This further described the text mining process steps undergone in extracting useful information from the customer survey feedbacks of one of the SUCs in the Philippines, the Bicol State College of Applied Sciences and Technology (BISCAST). Moreover, the Latent Dirichlet Allocation (LDA), a topic modeling method, was used for automatic text summarization and topic extraction from these text-based data. The topmost concerns extracted from the feedbacks were identified. This information provides useful insights for management analysis as well as inputs for policy making.