This research explores the novel use of the partial least squares regression (PLSR) as an alternative model to the conventional least squares (LS) model for modeling tide levels. The modeling is based on twenty tidal constituents: M2, S2, N2, K1, O1, MO3, MK3, MN4, M4, SN4, MS4, 2MN6, M6, 2MS6, S4, SK3, 2MK5, 2SM6, 3MK7, and M8. The 1st, 2nd, and 3rd PLSR components are selected from 40 PLSR components for the modeling based on the computed variances, Yloadings, and Yscores. The PLSR results are compared with those of the LS. The normality of the model residuals are evaluated by the Jarque–Bera statistical test. The computed probabilities of the normality test for 1st, 2nd, and 3rd PLSR components and LS are p = 0.0611, p = 0.0656, p = 0.916, and p = 0.0517, respectively, which all indicate p > 0.05, and imply that the residuals are normally distributed. The nature of tide criterion is verified by computing the tidal form factor F. The computed tidal form factor for the 1st, 2nd, and 3rd PLSR components and LS are F = 0.1794, F = 0.1696, F = 0.1599, and F = 0.1848 respectively. All the models satisfy the semidiurnal criterion of 0 ≤ F ≤ 0.25 at the 95% confidence level, since the observed tide is characteristically semidiurnal. The computed coefficient of determination for the 1st, 2nd, and 3rd PLSR components and LS are r² = 0.9134, r² = 0.9825, r² = 0.9933, and r² = 0.7861 respectively. These results prove that the PLSR model outperformed the conventional LS model, and therefore, a viable alternative to the conventional LS model.

中文翻译：

---

使用偏最小二乘回归进行潮汐建模

这项研究探索了偏最小二乘回归（PLSR）作为潮汐水位建模传统常规最小二乘（LS）模型的替代模型的新颖用法。该建模基于20个潮汐成分：M2，S2，N2，K1，O1，MO3，MK3，MN4，M4，SN4，MS4、2MN6，M6、2MS6，S4，SK3、2MK5、2SM6、3MK7和M8。基于计算出的方差，Yloadings和Yscores，从40个PLSR组件中选择第一，第二和第三PLSR组件。将PLSR结果与LS的结果进行比较。模型残差的正态性通过Jarque-Bera统计检验进行评估。第1，第2和第3 PLSR分量和LS的正态性检验的计算概率为p  = 0.0611，p  = 0.0656，p  = 0.916和p  = 0.0517，分别表示p  > 0.05，这意味着残差呈正态分布。通过计算潮汐形态因子F验证了潮汐判据的性质。为第一，第二，和第三PLSR组件和LS所计算潮汐形状因数是˚F  = 0.1794，˚F  = 0.1696，˚F  = 0.1599，和˚F 分别= 0.1848。所有的模型满足0≤半日标准 ˚F  ≤0.25在95％的置信水平，由于观测到的潮汐是典型半日。计算出的第一，第二和第三PLSR分量和LS的确定系数为r ²  = 0.9134，r²  = 0.9825，r ²  = 0.9933和r ²  = 0.7861。这些结果证明PLSR模型优于常规LS模型，因此是常规LS模型的可行替代方案。