若干电解质浓溶液活度系数估算的支持向量回归算法  

Support vector regression applied to estimation of activity coefficients of some concentrated electrolyte solutions

作　　者：张良苗[1] 陆文聪[1] 刘旭[1] 陈念贻[1] 姚莉秀[2] 

机构地区：[1]上海大学化学系计算机化学研究室,上海200436 [2]上海交通大学图象及模式识别研究所,上海200030

出　　处：《计算机与应用化学》2003年第6期745-749,共5页Computers and Applied Chemistry

基　　金：国家自然科学基金(50174038)

摘　　要：应用支持向量回归算法筛选出与电解质浓溶液活度系数相关的离子特性参数集:阴阳离子半径比(R_+/R_)、阴离子半径R_、阳离子半径R_+和阴阳离子电荷数比(Z_+/Z_)。并以此为自变量集,用支持向量回归算法或PLS算法总结活度系数的经验规律,进而提出利用一批浓电解质溶液已知的活度系数数据“转推”其他电解质溶液的活度系数的算法。用留一法考察了这种“转推”算法的准确程度。并引用离子系的对应态理论对支持向量回归求得的经验关系的物理意义作了讨论和解释。By using feature selection technique based on support vector regression, a set of features including ionic radius ratio (R+ /R- ) , cationic radius R+ , anionic radius R- and ionic charge ratio (Z+ /Z_ ) has been selected for computerized prediction of the activity coefficient of concentrated electrolytic solutions. Based on this feature set, some approximate linear relationship between the In r and the above-mentioned parameters has been found by support vector regression or PLS method. And a more accurate method for the estimation of the values of In r of the concentrated electrolytic solutions by support vector regression with kernel function of 2-degree has been proposed. The theory of corresponding state has been used to discuss the physical meaning of these computation methods

关 键 词：电解质浓溶液 活度系数 支持向量回归 计算方法 物理化学 

分 类 号：O646.1[理学—物理化学]