MOSCH: A MULTI-OBJECTIVE SPATIAL CLUSTERING ALGORITHM WITH CONSTRAINT-HANDLING METHODS
Baksa, Bridget Marie
Area of Honors:
Information Sciences and Technology
Bachelor of Science
Dr. Abdullah Konak, Thesis Supervisor Dr. Sandy Feinstein, Honors Advisor
Artificial Intelligence Machine Learning Data Science Clustering Algorithms Information Science Technology NYC CitiBike Big Data Software Engineering Python
Clustering algorithms are a popular machine learning technique used to classify data. While there are many variations of clustering algorithms, they each are specialized to perform one specific task. A multi-objective spatial clustering algorithm with constraint-handling methods (MOSCH) is proposed to combine the main aspects of popular machine-learning clustering algorithms. The proposed MOSCH algorithm calculates and minimizes the Euclidean Distance (E) and Mean Boundary Distance (MBD) of the data points within the data set. Since these two objectives are conflicting, MOSCH searches for a set of non-dominating solutions that meet both objectives to a certain degree. In addition, constraints are implemented to further customize the solution results. The clustering solutions were consistent with the predicted results for the test data set. MOSCH is also shown to be successful in clustering Citi Bike Stations in New York City to effectively manage bike transportation.