An activity improved robust adaptive normal SVR optimization algorithm

作者：He Feng; Sun Danfeng^*; Li Hong; Zhang Weiwei

来源：Journal of Computational Information Systems, 2014, 10(14): 6041-6047.

DOI：10.12733/jcis10993

• Convex combinations
• Evaluation modeling
• Least squares support vector regression
• Normal direction
• Optimization algorithms
• Shift directions
• Support vector regression algorithms
• Water quality evaluation

It is well-known that water quality evaluation is very important to protect water resource. This study is to propose robust adaptive normal support vector regression algorithm for water quality evaluation. Robust adaptive normal support vector regression algorithm can consider the normal direction of the output regressor as the shift direction of training samples, which can overcome this deficiency by considering the convex combination of training points with the smallest projection values. As dissolved oxygen is the important index to measure the water quality, we can understand water quality by predicting dissolved oxygen in the water. The experimental results indicate that mean error of dissolved oxygen prediction of robust adaptive normal support vector regression algorithm is 0.0115, mean error of dissolved oxygen prediction of least squares support vector regression algorithm is 0.0240, and mean error of dissolved oxygen prediction of traditional support vector regression algorithm is 0.0545. It can be seen that robust adaptive normal support vector regression algorithm is more excellent dissolved oxygen prediction ability than least squares support vector regression algorithm or traditional support vector regression algorithm. Thus, we can conclude that robust adaptive normal support vector regression algorithm is very suitable for water quality evaluation.

• 出版日期2014
• 单位中国农业大学