THE EVALUATION SYSTEM OF DESIGN SOLUTIONS FOR RESIDENTIAL PROPERTY ON THE PRE-INVESTMENT STAGE THROUGH NEURAL NETWORK TECHNOLOGY

Abstract

Ever since the USSR timeline design solutions have been assessed according to various criteria and indicators. Systematic and technical doctrine used for evaluation of design solutions distinguishes itself at the present stage. The doctrine has been complemented by theory of efficiency and financial sustainability of an investment project with due account of general market concept. Also great attention is paid to the virtual object modeling. It looks rather important to include behavior prediction of an investment construction project model at each stage of its life cycle. High cost of all life cycle stages pertainning to the investment construction project stipulates the necessity to calculate an investment prior to design works while assessing an investment idea. Pre-investment stage of construction project development has been secured in legislation of the Republic of Belarus. In order to evaluate a design solution at this stage it is necessary to develop an investment justification, a project management plan and a business plan which make it possible fully assess and compare several variants of future objects. such approach requires not only time, but considerable financial costs. It has been proposed to develop an evaluation system for design solutions for optimization of the given process and the system is based on the existing projects. The system allows a customer (an investor) to select a variant of design solution excluding development of pre-design documentations for several options while constructing an immovable property object. It is expediently to test the given system while using construction of multifamily residential building with involvement of Republican collection of design documentation and objects-analogues databank as an example. It is presupposed to use a theory of neural networks and neuro-programming in the developed system for assessment of design solutions for residential real estate objects at a pre-investment stage. This system bases on the input parameters projects. While using the system trained neurons of hidden layer select suitable projects of multifamily residential buildings on the basis of input parameters apartment houses with their classification. The projects are classified in decreasing order of summary significance of main output parameters. As a result, a customer obtains predicted main parameters of the future investment project without development of pre-design documentation package.