To find earthquake precursory items that are closely relative to earthquake is very important for earthquake prediction. Rough Set Theory is an important tool to process imprecise and ambiguous information. In this paper, the discernibility matrix approach based on Rough Set Theory is optimized to find all possible rough reducts with reduced time and space complexity. Furthermore, this approach is applied to analyze the dependency among earthquake precursory items. After several experiments, some most important precursory items are found, while some items are considered to be redundant. The results maybe provide the researches with the direction on the relationship between earthquake precursory items and earthquake.