How does factor network analysis differ from conventional regression analysis and neural networks?
Regression analysis assumes that the inputs are totally independent, which is never the case for historical process data. As a result, regression analysis can lead to incorrect or unstable results when analyzing large amounts of interrelated data. While neural networks use a structure similar to factor network analysis, neural networks also suffer from similar limitations when dealing with interrelated input data. However, the largest difference is FactNet’s ability to combine statistics with the user’s process experience. No other type of analysis, statistical or otherwise, allows qualitative process experience to be merged with rigorous statistical methods.
