Traditional chemical process monitoring methods focus on higher detection rate and ignore monitoring system cost. To minimize the cost, we propose improved cost-optimal Bayesian control chart for auto-correlated chemical process monitoring. First, the least square support vector machine (LSSVM) is used to model the auto-correlated process and obtain the independent residuals. Then, a two-condition hidden Markov model (HMM) is used to describe the residuals. Finally, the cost-optimal Bayesian control chart is developed through semi-Markov decision process (SMDP) framework to achieve cost-optimal control limit minimizing the long run expected average cost as well as the monitoring statistic. The monitoring results verify that the improved cost-optimal Bayesian control chart achieves better economic performance. (C) 2017 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.