A Quasi-Alternating Markov-Modulated Linear Regression: Model Implementation Using Data about Coaches’ Delay Time

Nadezda Spiridovska

International journal of circuits, systems and signal processing (1998-4464), Vol. 12, pp. 617-628(2018)
Keywords: External environment, delay time analysis, Markov-modulated linear regression, trip time

Open access: DOWNLOAD FULL TEXT  Export citation: BibTeX RIS

Abstract:

This research presents a case-study of quasialternating Markov-modulated linear regression application for analysis of delays of coaches (regional buses) on the route VentspilsRiga in Latvia. Markov-modulated linear regression suggests that the parameters of regression model vary randomly in accordance with external environment which is described as a continuous-time homogeneous irreducible Markov chain with known parameters. Markov-modulated linear regression model differs from other switching models by a new analytical approach. For each state of the environment the regression model parameters are estimated. External environment has only two states in this research that is why model is called quasi-alternating. Data on weather conditions provided by the Latvian Environment, Geology and Meteorology Centre and is free downloaded from its database. Data on weather conditions in the Ventspils city are used for the environment description: two alternate states are assumed: “no precipitation” and “precipitation”. The model of the external environment is tested for the markovian properties using inferential statistics. Actual data on coaches’ trip times is provided by the Riga International Coach Terminal. Data are analysed by means of descriptive statistics. Different experiments are carried out and the application of Markov-modulated linear regression model on given sample showed adequate results indicating the validity of the model