Estimation Method Based on MinMaxEnt Distribution for Missing Value in Time Series

Shamilov A., Giriftinoglu C.

8th WSEAS International Conference on System Science and Simulation in Engineering, Genoa, Italy, 17 - 19 October 2009, pp.206-210 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • City: Genoa
  • Country: Italy
  • Page Numbers: pp.206-210
  • Keywords: MinMaxEnt Distribution, MaxEnt Distribution, Time Series with Missing Value, Estimation of Missing value, Autoregressive Modelling
  • Anadolu University Affiliated: Yes

Abstract

In this study, MinMaxEnt distribution out of Generalized Entropy Optimization Distributions with moment functions and moment values dependent on parameter are introduced. MaxEnt distribution for observed time series is obtained as a multivariate normal distribution. Then MinMaxEnt distribution for time series with missing value is constituted and the value gamma determined by MinMaxEnt distribution is a estimation for the missing value in times series. Estimation method based on MinMaxEnt distribution for missing value is applied for the time series generated by AR(4) process. By assumption that the value in each position in time series is missing, mean square error (MSE) is calculated for the mentioned new estimation method.