零膨脹計數資料為實務分析中常見的資料型態。過往文獻多採用零膨脹卜瓦松（Zero-Inflated Poisson, ZIP）模型進行分析，透過機率混合的方式處理過多零值的問題。有別於此類方法，本研究探討閾值邊界卜瓦松迴歸（Threshold Boundary Poisson Regression, TBPR）模型於零膨脹計數資料之估計問題。TBPR 模型透過閾值邊界函數將樣本劃分為兩個子群，並於各子群分別配適獨立之卜瓦松迴歸模型。為估計 TBPR 模型，本研究提出一種結合加權支持向量機（Weighted Support Vector Machine, WSVM）與 L2 正則化卜瓦松迴歸之迭代演算法，稱為 TBPR-WSVM 演算法。該演算法利用初始分群建立子群，再藉由加權支持向量機更新閾值邊界並重新劃分樣本，並於每次迭代中採用 L2 正則化估計各子群之卜瓦松迴歸係數。模擬結果顯示，TBPR-WSVM 演算法具有較穩定的預測表現；於實證資料分析中，相較於 ZIP 模型，亦展現較佳之預測能力。

Zero-inflated count data are commonly encountered in applied analysis. The literature has predominantly relied on the zero-inflated Poisson (ZIP) model for such data, which addresses the excess of zeros through probability mixing. Departing from this approach, this study investigates the estimation of the Threshold Boundary Poisson Regression (TBPR) model for zero-inflated count data. The TBPR model partitions samples into two subgroups via a threshold boundary function and fits an independent Poisson regression on each subgroup. To estimate the TBPR model, this study proposes an iterative algorithm that combines a Weighted Support Vector Machine (WSVM) with L2 regularized Poisson regression, referred to as the TBPR-WSVM algorithm. The algorithm first constructs subgroups through an initial partitioning step, then updates the threshold boundary and reclassifies the samples via a weighted support vector machine, and estimates the Poisson regression coefficients within each subgroup using L2 regularization in each iteration. Simulation results show that the TBPR-WSVM algorithm achieves more stable predictive performance. In real data analysis, it also demonstrates better predictive capability than the ZIP model.

Boyd, S., & Vandenberghe, L. (2004). Convex optimization. Cambridge University Press.
Brooks, M. E., Kristensen, K., van Benthem, K. J., Magnusson, A., Berg, C. W., Nielsen, A., Skaug, H. J., Mächler, M., & Bolker, B. M. (2017). glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. The R Journal, 9(2), 378–400. https://journal.r-project.org/articles/RJ-2017-066/
Chang, C.-H., Emura, T., & Huang, S.-F. (2026). An algorithm for estimating threshold boundary regression models. Computational Statistics & Data Analysis, 108274.
Chen, C.-X. (2024). Estimation of threshold boundary Poisson regression models [Master’s thesis]. National University of Kaohsiung.
Chen, Y.-L. (2023). Estimation of threshold boundary logistic regression models [Master’s thesis]. National University of Kaohsiung.
Cragg, J. G. (1971). Some statistical models for limited dependent variables with application to the demand for durable goods. Econometrica, 39(5), 829–844.
Farebrother, R. W. (1976). Further results on the mean square error of ridge regression. Journal of the Royal Statistical Society: Series B (Methodological), 38(3), 248–250.
Fong, Y., Huang,Y.,Gilbert, P. B., &Permar,S.R (2017). chngpt:Threshold regression model estimation and inference. BMC Bioinformatics, 18(1), 454.
Hansen, B. E. (2000). Sample splitting and threshold estimation. Econometrica, 68(3), 575–603.
Hoerl, A. E., & Kennard, R. W. (1970). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics, 12(1), 55–67.
Köll, S., Kosmidis, I., Kleiber, C., & Zeileis, A. (2021). Bias reduction as a remedy to the consequencesof infinite estimates in Poisson and Tobit regression. arXivpreprintarXiv:2101.07141.
Lambert, D. (1992). Zero-inflated Poisson regression, with an application to defects in manufacturing. Technometrics, 34(1), 1–14.
le Cessie, S., & van Houwelingen, J. C. (1992). Ridge estimators in logistic regression. Journal of the Royal Statistical Society: Series C (Applied Statistics), 41(1), 191–201.
Lee, S., Liao, Y., Seo, M. H., & Shin, Y. (2021). Factor-driven two-regime regression. The Annals of Statistics, 49(3), 1656–1678.
Månsson, K., & Shukur, G. (2011). A Poisson ridge regression estimator. Economic Modelling, 28(4), 1475–1481.
Mullahy, J. (1986). Specification and testing of some modified count data models. Journal of Econometrics, 33(3), 341–365.
Quandt, R. E. (1958). The estimation of the parameters of a linear regression system obeying two separate regimes. Journal of the American Statistical Association, 53(284), 873–880.
Roulin, A., & Bersier, L.-F. (2007). Nestling barn owls beg more intensely in the presence of their mother than in the presence of their father. Animal Behaviour, 74(4), 1099–1106.
Schaefer, R. L., Roi, L. D., & Wolfe, R. A. (1984). A ridge logistic estimator. Communications in Statistics– Theory and Methods, 13(1), 99–113.
Seo, M. H., & Linton, O. (2007). A smoothed least squares estimator for threshold regression models. Journal of Econometrics, 141(2), 704–735.
Silva, J. M. C. S., & Tenreyro, S. (2010). On the existence of the maximum likelihood estimates in Poisson regression. Economics Letters, 107(2), 310–312.
Theobald, C. M. (1974). Generalizations of mean square error applied to ridge regression. Journal of the Royal Statistical Society: Series B (Methodological), 36(1), 103–106.
Tong, H., &Lim,K.S.(1980).Threshold autoregression, limit cycles and cyclical data. Journal of the Royal Statistical Society: Series B (Methodological), 42(3), 245–268.
Yang, L., Ren, M., & Bai, J. (2025). Threshold mixed data sampling logit model with an application to forecasting US bank failures. Empirical Economics, 68(1), 433–477.
Yu, P., & Fan, X. (2021). Threshold regression with a threshold boundary. Journal of Business & Economic Statistics, 39(4), 953–971.