A polynomial regression method based on Trans-dimensional Markov Chain Monte Carlo

In this paper, we focus on the problem of polynomial regression with uncertain polynomial order and coefficients. For polynomial regression problem, a probabilistic graph model with polynomial coefficients and polynomial order is established firstly. Then Bayesian inference is conducted with Trans-dimensional Markov Chain Monte Carlo (TDMCMC) sampling approach to achieve a Bayesian polynomial regression method, which can self-adaptively determine polynomial parameters and order with trans-dimensional strategy. The performance of the developed algorithm is demonstrated via true polynomial regression experiment. The experimental results show that the proposed method is effective.