1: \begin{abstract}

2: We consider to model matrix time series based on a tensor CP-decomposition.

3: Instead of using an iterative algorithm which is the standard practice for estimating CP-decompositions, we propose a new and

4: one-pass estimation procedure based on a generalized eigenanalysis constructed

5: from the serial dependence structure of the underlying process.

6: To overcome the intricacy of solving a rank-reduced generalized eigenequation, we propose a further refined approach which projects it into a lower-dimensional full-ranked

7: eigenequation. This refined method improves significantly the finite-sample performance of the estimation.

8: % {\color{red}A key idea of the new procedure is to project a generalized eigenequation defined in terms of rank-reduced matrices to a lower-dimensional one with full-ranked matrices, to avoid the intricacy of the former of which the number of eigenvalues can be zero, finite and infinity.}

9: The asymptotic theory has been established under a general setting without the stationarity. It shows, for example, that all the component coefficient vectors in the

10: CP-decomposition are estimated consistently with certain convergence rates. 

11: %depending on the relative sizes between the dimensions of time series and

12: %the sample size. 

13: The proposed model and the estimation method are also illustrated with both simulated and real data;

14: showing effective dimension-reduction in modelling and forecasting matrix time series.

15: \end{abstract}

16: