Conditions of Application of the Modified Least Square Method for Models Identification of Distributed Dynamic Systems

The identification of the parameters of distributed dynamical systems is an important task for technical, economic and social applications. The quality of statistical estimates is usually determined by two criteria: standard error and bias. But most of the methods cause the inconsistency of the criteria. The desire to reduce the standard error generates an increase in bias and vice versa, such a problem is well known in machine learning. The paper proposes a modified least squares method (OLS), the use of which provides a reduction in both criteria. The conditions for its application are considered.