In recent years, social media platforms have become invaluable sources of information, providing a real-time reflection of public opinions, and discussions. Among these platforms, Twitter has emerged as a dynamic space where individuals share their thoughts on various subjects, including their experiences with airlines. Manual analysis of tweets about Ethiopian Airlines is time-consuming and inefficient. There is a need for an automated approach to categorize tweets into topics, allowing for timely response and actionable insights. The main objective of this research is, detecting topics discussed on twitter about the Ethiopian airlines. In this paper, latent dirichlet allocation-variation expectation maximization, latent dirichlet allocation- gibbs sampling, and correlated topic model-variation expectation Maximization have been used for topic detection. Four distinct topic detection tests have been carried out by varying the topic number and seed value of the algorithms in an effort to identify a model that produces more cohesive and useful topics. Using the same datasets, we compare three topic detection techniques. The models' performance from these four experiments is analyzed and assessed both qualitatively (manually) and quantitatively (entropy). We observed how the type of topic model, the number of topic and seed value affect the quality of detected topics, Among the three models, the latent dirichlet allocation-variation expectation maximization model with K = 6 and Seed value = 1000 has shown the lowest entropy selected as the best model. The most topics discussed about Ethiopian airlines ―customer‖, ― services‖, ―lost‖, ― luggage‖ and ―canceled‖, ― flight‖ The results from this study were encouraging; it is the researcher‘s belief that a more detailed study of tweets using data mining techniques and managing this information obtained from customers can help Ethiopian airlines to adjust their services to perfectly fit customer need. Keywords: Topic detection, Latent Dirichlet allocation, Gibb’s sampling, Variation Expectation Maximization, Topic modeling, API, social media