Data Generation using Large Language Models for Text Classification: An Empirical Case Study