Improving of the Identification Algorithm for a Quasilinear Recurrence Equation

Identification of quasilinear recurrence equations (QRE) may be reduced to the problem of regression analysis with mutually dependent observable variables. It is possible to use the generalized least deviations method (GLDM) for such problems. GLDM-estimation consists of solving the sequence of the WLDM-estimation problems. We propose the algorithm to solve the WLDM-estimation problem. Computational complexity of this algorithm does not exceed the quantity \(O(N^2T+T^2)\), where N is the number of coefficients in the considered equation, T is the number of observed readings. The computational complexity of solving practical GLDM estimation problems does not exceed \(O(N^3T+NT^2)\). Results of computational experiments to solve the problem of identifying the recurrence equation of the stock market index in Iraq by original data from the site “ISX-IQ.net” are presented. This results show the possibility to apply a second order quasilinear recurrence equation with quadratic nonlinearity for these purposes. Perhaps increasing the order of the recurrence equation and the accuracy of the calculations give better results.