摘要
利用数据挖掘方法对1990年1月—2001年7月10余年的WCPO金枪鱼围网作业生产数据以及该时段的18个与WCPO金枪鱼围网产量相关的海洋环境因子进行分析,将信息增益分析技术和基于粗糙集的属性化简方法相结合用以确定影响WCPO金枪鱼围网产量的关键因子集。利用多元回归分析方法建立基于关键因子集的预测模型,经过大量试验对比选择较佳的预测模型,达到了理想的效果。同时,验证了所确定关键因子集的可靠性,有效性。首次利用多种数据挖掘方法相结合的方式对多种影响因子进行分析,开展渔情预报研究,达到了较好的效果,为渔情预报研究提供了一种新思路,同时为渔业生产提供了可靠的依据。
In order to find the key influence attributes set to tuna purse seiner output in WCPO, the attained data of tuna purse seiner output from Jan 1990 to Jul 2001, and eighteen environmental factors associated with tuna purse seiner output in the same period were analyzed with the information gain analysis method and rough set theory. Several fishery prediction models were established by multiple regression analysis based on the key attributes set and then were compared to confirm the best one. The good prediction effect of that model proved that the key attribute set contains the main information of those influence attribute data. Two data mining methods were combined to analyze multiple factors, which could be a new analyzing method of fishery prediction study. The prediction model based on the key influence attributes set improves the fishery prediction precision and provide the important information to fishery industry.
出处
《海洋技术》
北大核心
2012年第1期103-106,共4页
Ocean Technology
基金
国家高技术研究发展计划(863)资助项目(2003AA637030)
关键词
中西太平洋
金枪鱼围网
信息增益
粗糙集
渔情预报
western and central Pacific Ocean tuna purse seine fishery information gain rough set fishery forecast
