Describir: Prediction of silicon content in hot metal based on golden sine particle swarm optimization and random forest