Learning hidden Markov models with persistent states by penalizing jumps
作者：
Highlights：
• Estimate hidden Markov models by clustering temporal features while penalizing jumps.
• Learn the hidden state sequence and model parameters simultaneously and faster.
• Control the transition rate to avoid unrealistically rapid switching dynamics.
• Compares favorably to spectral clustering and maximizing the likelihood function.
• Improved estimates result in lower transaction costs when used in a trading strategy.
摘要
•Estimate hidden Markov models by clustering temporal features while penalizing jumps.•Learn the hidden state sequence and model parameters simultaneously and faster.•Control the transition rate to avoid unrealistically rapid switching dynamics.•Compares favorably to spectral clustering and maximizing the likelihood function.•Improved estimates result in lower transaction costs when used in a trading strategy.
论文关键词：Time series analysis,Clustering,Unsupervised learning,Regime switching,Regularization,Dynamic programming
论文评审过程：Received 13 August 2019, Revised 22 January 2020, Accepted 11 February 2020, Available online 13 February 2020, Version of Record 21 February 2020.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.113307
