School of Water Conservancy & Civil Engineering / Northeast Agricultural University, Harbin, China

---

Abstract—In order to improve fitting and forecasting precision and solve the problem that some data with less sensitivity lead to low simulation precision of partial leastsquares regression (PLS for short) model, the new method of simulating freezing depth is presented according to ground temperature of different depths, air temperatures, surface temperatures and the like. Firstly, the PLS model, which is built by virtue of the ideas of principal component analysis and canonical correlation analysis, can be adopted to solve the multi-correlation among each factor effectively by extracting principal components. And the interpretation ability of each principal component to freezing depth can be obtained by assistant analysis. Meanwhile, particle swarm optimization algorithm (PSO) is adopted to optimize partial regression coefficient, and then the PLS model based on PSO can be built. Compared with traditional PLS model, the optimized model has more reliability and stability, and higher precision.

Index Terms—freezing depth, particle swarm optimization, simulation, partial least-squares regression, principle component

[PDF]

Cite: Tianxiao Li, Qiang Fu, Fanxiang Meng, Zilong Wang, and Xiaowei Wang, "Study on Partial Least-Squares Regression Model of Simulating Freezing Depth Based on Particle Swarm Optimization," Journal of Computers vol. 6, no. 3, pp. 532-539, 2011.