研究工业酶非分类关系抽取问题。针对传统关联规则抽取非分类关系时存在效率低、收敛速度慢及漏报规则等问题，提出了一种基于小生境技术的萤火虫算法。该算法利用小生境技术的融合、演化算法丰富种群的多样性，结合萤火虫算法寻优速度快的优势抽取非分类关系，解决了局部最优、规则冗余问题。针对工业酶语料进行了验证性实验，实验结果表明，相对于传统的关联规则挖掘算法，该方法在个体多样性及提取有效规则的效率上都有较大的提高，挖掘结果对工业酶非分类关系抽取具有一定的参考价值。%This thesis focuses on the extraction of industrial enzyme non-taxonomical relations. Concerning the shortage of low efficiency, slow convergence and omission rules caused by traditional methods, the firefly algorithm which is based on niche technology is proposed. The method solves the problem of algorithm premature and redun-dancy rules, using the integration of niche technology, population diversity of the algorithm and the non-taxonomical relations of the firefly algorithm. The experiment is carried out based on the industrial enzyme corpus. Compared with traditional association rules mining method, the results show that the algorithm performs better in terms of pop-ulation diversity and efficiency of discovering more association rules. The mining result offers reference value in in-dustrial enzyme non-taxonomic relation extraction.
展开▼
