东北林业大学学报
東北林業大學學報
동북임업대학학보
JOURNAL OF NORTHEAST FORESTRY UNIVERSITY
2014年
9期
39-42,57
,共5页
复共线性%岭估计%地类%蓄积量估测
複共線性%嶺估計%地類%蓄積量估測
복공선성%령고계%지류%축적량고측
Multicollinearity%Ridge estimation%Land category%Forest volume estimation
为克服自变量间多重相关性对森林蓄积量估测的不利影响,采用岭估计建立估测模型。在确定岭参数时,分别试验了岭迹法、方差扩大因子法、Hoerl-Kennard法和预测残差平方和法,对不同地类分别建立蓄积量估测方程。根据建模样地的蓄积量拟合残差和预报样地的蓄积量估测偏差分别计算中误差,利用残差中误差(δ^R)和预报偏差中误差(^δF )构造评价函数0.5(^δR+^δF ),并测度岭参数的优劣,进而确定监测区域的估测模型。结果表明:当影响蓄积量估测的自变量间存在复共线性时,分类建立模型的精度明显优于统一建模的精度。
為剋服自變量間多重相關性對森林蓄積量估測的不利影響,採用嶺估計建立估測模型。在確定嶺參數時,分彆試驗瞭嶺跡法、方差擴大因子法、Hoerl-Kennard法和預測殘差平方和法,對不同地類分彆建立蓄積量估測方程。根據建模樣地的蓄積量擬閤殘差和預報樣地的蓄積量估測偏差分彆計算中誤差,利用殘差中誤差(δ^R)和預報偏差中誤差(^δF )構造評價函數0.5(^δR+^δF ),併測度嶺參數的優劣,進而確定鑑測區域的估測模型。結果錶明:噹影響蓄積量估測的自變量間存在複共線性時,分類建立模型的精度明顯優于統一建模的精度。
위극복자변량간다중상관성대삼림축적량고측적불리영향,채용령고계건립고측모형。재학정령삼수시,분별시험료령적법、방차확대인자법、Hoerl-Kennard법화예측잔차평방화법,대불동지류분별건립축적량고측방정。근거건모양지적축적량의합잔차화예보양지적축적량고측편차분별계산중오차,이용잔차중오차(δ^R)화예보편차중오차(^δF )구조평개함수0.5(^δR+^δF ),병측도령삼수적우렬,진이학정감측구역적고측모형。결과표명:당영향축적량고측적자변량간존재복공선성시,분류건립모형적정도명현우우통일건모적정도。
For overcoming adverse effects of the multiple correlations between the independent variables on forest stock volume estimation, we used ridge estimation to establish estimation models.In determining ridge parameters, we respectively tested ridge trace method, variance inflation factor method, Hoerl-Kennard method and predictive residual sum of squares, and established stock volume estimating equations for different land types.According to the stock volume fitting residue of mod-eling plots and the stock volume estimation bias of prediction plots, we calculated mean square error, construct evaluation function by using the residual error and the prediction error, measured the pros and cons of ridge parameters, and deter-mined the estimation model of monitoring area.When there is multicollinearity between independent variables that affect stock volume estimation, the precision of classification modeling will be significantly better than unified modeling.