The Application of the Neuro-Fuzzy Computing Technique for the Forecasting of the R&D Project Performance
Yu‐Shan Chen, Ke‐Chiun Chang · 2006
This study used the adaptive neuro-fuzzy inference system (ANFIS) and ordinary least squares (OLS) regression to forecast the R&D project performances of Taiwanese IC design companies through three explanatory variables: the fitness of project environment, R&D project manager's skills, and the effectiveness of team work. The results showed that the accuracy rate of ANFIS in this study was 65.52% better than 55.17% of OLS regression model. Therefore, the ANFIS is more accurate than OLS regression to forecast the R&D project performance. Besides, this paper investigated the relationships between the R&D project performance and its determinants, and pointed out their nonlinear nature under the complex and uncertainty environment nowadays. This study showed that these three explanatory variables had inverse U-shaped effects on the R&D project performance with ANFIS which had more managerial implications than OLS regression which only indicated that these three explanatory variables were positively associated with the R&D project performance. Hence there existed optimal levels and U-shaped effects of these three determinants for the R&D project performance