Combining traditional ML with LLM-generated synthetic data: A balanced approach for text classification