Feature selection for high-dimensional data: a fast correlation-based filter solution