Analytic Scoring of Landscape CAD Course Based on BP Neural Network

Abstract

Course scoring is the major criteria for measuring the teaching effect and a basis for improving the education quality. Analytic scoring of landscape CAD (computer-aided design) course is influenced by various factors. Relationships among these factors are complex, some are nonlinear, even some are random and fuzzy. It is difficult to explain their internal relationships with traditional method. This research combines back-propagation neural network and DPS software to establish a three-layer BP neural network model, which took 60 examination papers of landscape CAD course as samples and made predictions on the score in accordance with five factors, including landscape design standard, landscape design innovation, computer cartography standard, drawing effect and workload. The results show that BP neural network model has strong nonlinear approximation ability, could truly reflects the nonlinear relationships between global score of landscape CAD course and main controlling factors of analytic scoring, with small error between predicted values and the measured values, relative error lower than 5%. In the future, when analytic scoring of the landscape CAD course obtains from the teachers, the global scoring can be calculated by BPNN model automatically. This method showed wide application prospect to the courses need analytic scoring.