ANCFIS-ELM: A machine learning algorithm based on complex fuzzy sets
Omolbanin Yazdanbakhsh, Scott Dick · 2016
The Adaptive Neuro-Complex Fuzzy Inferential System was the first neuro-fuzzy system employing complex fuzzy sets and rule interference. It was shown to be both accurate and parsimonious in time series forecasting. The main disadvantage of this system is its slow learning algorithm. One possible approach to speeding up this neuro-fuzzy system is to apply concepts from the Extreme Learning Machine family of architectures; specifically, we will randomly select the parameters of a “pool” of complex fuzzy sets, and then train the neural network by incrementally updating the parameters of a linear output function. We evaluate this new architecture on four software reliability growth datasets (a particular instance of time series forecasting).