Extreme learning with Artifical Electric Field Algorithm (EL-AEFA) for estimation of Short-term Vehicular Traffic System
Sarat Chandra Nayak, Sanjib Kumar Nayak, Alina Dash · 2021 IEEE 8th Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON) · 2021
The faster learning and better approximation ability of extreme learning machine (ELM) made it a popular learning method for neural network training. However, two issues such as probable inclusion of non-optimal parameters because of random selection and size of hidden layer may unfavorably impact the network performance. To address these issues, this article combines the artificial electric field algorithm (AEFA) and ELM to construct a new learning framework called as extreme learning with AEFA (EL-AEFA). Then, EL-AEFA is used to find the optimal model parameters (i.e., weight, bias and number of hidden neurons) of a neural network with single hidden layer. The proposed method is used to estimate the short-term traffic volume through exploitation of data available publicly. The model exhibits minimal structural and computational complexity, follows adaptive training and generates lowest mean absolute percentage of error (MAPE) values compared to few other forecasts developed in a similar way. Therefore, it may be recommended as a promising tool for traffic volume prediction.