Key words
linear-multiplicative regression, least absolute deviation, partially Boolean linear programming, program complex, volume of passage of large-capacity containers, Zabaikalsk-Manchuria railway crossing point.
articles
The author:
Bazilevskiy M. Published in
№ 3(75)
29 june 2018 year
Rubric: Models and Methods
Rubric: Models and Methods
Program complex for constructing the linear-multiplicative regressions
When constructing regression models, the key stage is the model specification, which assumes the
choice of composition and the mathematical form of the relationship between the variables in regression
equation. To date, there is no system of standard recommendations and methods that would form
a rigorous theoretical basis for selecting a model specification. The article is devoted to the problem
of specification of regression models, namely, the subset selection in linear-multiplicative regressions.
Linear-multiplicative regressions are non-linear in factors, but linear in parameters, and reflect the degree
of joint influence of independent variables on the dependent variable. This problem can be formalized
as a problem of partial-Boolean linear programming. Since the solution of such problems requires
the computational capabilities of modern computers, a universal software package was developed
for constructing linear-multiplicative regressions, which can be used in technical studies, economics,
business, sociology, medicine, etc. To demonstrate its work, the problem of volume modeling
passing large-capacity containers at the Zabaikalsk-Manchuria railway checkpoint. At the same time,
the speed of solving such computational problems was tested.
The author:
Bazilevskiy M.
Degree:
PhD in Technique, Associate Professor, Irkutsk State Transport University
Location:
Irkutsk
