This research builds Internet growth forecasting models based on existing knowledge of diffusion and connectionist theories. It shows that a simple connectionist multi-layered perceptron artificial neural network (MLP) model can create a flexible response function to forecast Internet growth for the near future. This paper identifies the most suitable diffusion models that generate predictions for the Internet diffusion with low errors. However, the MLP model is superior to the best diffusion model on both the calibration and the validation samples of Internet growth data. This research also investigates the process of combining diffusion and connectionist models. The findings will encourage researchers to use connectionist models to predict diffusion of other innovation processes also.
"Global Diffusion of the Internet IX: Predicting Global Diffusion of the Internet: An Alternative to Diffusion Models,"
Communications of the Association for Information Systems: Vol. 17
, Article 5.
Available at: http://aisel.aisnet.org/cais/vol17/iss1/5