POP-TRAFFIC: A Novel Fuzzy Neural Approach to Road Traffic Analysis and Prediction

GA-Based Learning Algorithms to Identify Fuzzy Rules for Fuzzy Neural Networks

“…The structure of the fuzzy neural network model used in this paper is similar to the structure proposed in [2]. It is a fivelayer structure, as shown in Figure 1, where each layer performs an operation for building the fuzzy system.…”

“…First stage is initializing the membership functions of both input and output variables by determining their centres and widths. To perform this stage, we have employed a self-organizing algorithm [6] as in other works [2,5,16]. A proposed GA based learning algorithm is performed in the second stage to identify the fuzzy rules that are supported by the set of training data.…”

“…An important topic in designing of a fuzzy neural network is the identification of the fuzzy rules. However, there is no systematic design procedure at present [2]. The recent research direction in the identification of the fuzzy rules in fuzzy neural networks is to learn and modify the rules from past experience.…”

“…This paper extends this idea further by incorporating a weight parameter to trade-off between the number of fuzzy rules and overall error. In this paper we have used GA to make use of the known membership function to identify the fuzzy rules using similar fuzzy neural network structure discussed in [2]. The proposed genetic algorithm differs form existing algorithms such as [5,6], in terms of being simple, fast and flexible to control the process of identification of the fuzzy rules based on the error level.…”

Analysis of Large-Scale Traffic Dynamics in an Urban Transportation Network Using Non-Negative Tensor Factorization

“…In this branch, Kalman filter [14] and ARMA (Autoregressive Moving Average) [15], originated from state space theory, are popularly used to predict linear variation tendency of traffic flows [14][15] [19]. In [20][21] [22], neural networks [20] [21] and hybrid non-linear dynamic systems [22] are used to approximate short-term non-linear fluctuations of traffic flow states. Due to the intrinsic multiple-input and multiple-output (MIMO) structures, neural networks intrinsically integrate spatio-temporal correlations between local link segments.…”