摘要
赤潮的危害日益严重,为了预测赤潮的发生,运用回归树的随机梯度Boosting算法分析渤海赤潮数据,建立浮游植物总量与环境因子的定量关系,给出各种环境因子对浮游植物总量相对影响的大小以及浮游植物总量和各种环境因子偏相关的图形,有利于探索赤潮的发生机制,指导菌种的培养。最后,相比其它算法,回归树的随机梯度Boosting对于“局部剧增”的赤潮数据是稳健的,而且具有较高的预测精度。
It is of great importance to predict occurrence of HAB since its harmful results are getting more and more serious. Moreover, functional relation, Relevant effect and partial dependence graphics between phytoplankton and each environmental factor can be concluded from survey data about HAB in Baohai by Stochastic Gradient Boosting of Regression Tree, which is essential for the research of cause of HAB and instruction of incubation of thalli. It is showing that stochastic gradient boosting which have better performance than support vector regression, artifical neural network, MARS and project pursuit regression is more robust to HAB data which is locally sharply increased.
出处
《海洋技术》
2005年第3期65-69,共5页
Ocean Technology
基金
国家自然科学基金资助项目(104723077)
