Estimating Alexa Rank Logic Using Fuzzy Adaptive Neuro-Fuzzy Inference System (ANFIS)

The aim of this study is to model the logic governing the rankings obtained from the Alexa ranking engine using available indicators through data fitting.

Six criteria, as input indicators for 233 top Iranian websites along with their Alexa rank as the output, constitute the general model of the problem. To implement this model, the Fuzzy Adaptive Neuro-Fuzzy Inference System (ANFIS) is trained on this data and finally a fitting with the least error is confirmed.

In this research, by comparing ANFIS models with different membership functions and different outputs through two methods of network partitioning and reduction clustering - considering the minimum errors generated - ANFIS with Gaussian membership functions and linear output applied by the network partitioning method had the best performance in estimating website rankings.

A relatively clear picture of the influence of various indicators and the estimation of future website ranks was presented by providing new or modified values of the desired indicators so that ultimately, using indicators other than the exact site traffic, the next rank could be estimated.