List of Articles multilayer perceptron network

• Open Access Article
  • Abstract Page
  • Full-Text
  
  1 - Short-term prediction of carbon monoxide gas concentration in the air of Ahvaz city using artificial neural network analysis
  Maryam Kavosi سیما سبزعلی پور hossein fathian
  Introduction: Air pollution in cities is one of the most critical environmental problems, representing a constant and severe threat to both the health and hygiene of society and the environment. The primary air pollutants include nitrogen oxides, with a particular empha More

  Introduction: Air pollution in cities is one of the most critical environmental problems, representing a constant and severe threat to both the health and hygiene of society and the environment. The primary air pollutants include nitrogen oxides, with a particular emphasis on nitrogen dioxide, sulfur oxides, especially sulfur dioxide, hydrocarbons, carbon monoxide (CO), carbon dioxide, and suspended particles. Ahvaz, a metropolis in Iran, stands out as one of the most polluted cities. Effective environmental management, particularly in addressing air pollution, is of paramount importance. This research aims to predict the concentration of CO pollutants in Ahvaz city for the first seven days of 2015. Materials and Methods: Based on previous studies, meteorological variables including weather, air temperature and wind speed were selected as gas input titles in the network for gas prediction. CO gas was procured in 2014 through the Environmental Protection Organization of Ahvaz city. In order to develop the Multilayer Perceptron (MLP) neural network, Neuro Solution5 software was used to create the neural network, 70% of the data was used for training (validation), 15% for testing, and the remaining 15% for validating the results of the network. is used. was used. Results and Discussion: In order to determine the best MLP network structure for short-term prediction of CO gas concentration, different structures were considered in terms of the number of intermediate layers, the type of network training algorithm, the type of transfer function, the number of intermediate layer neurons and the number of repetitions (Epoch) of training. The results showed that the MLP network with a structure of 1-5-3 (that is, 3 input neurons, 5 neurons in the middle layer and one neuron for the output layer) with 1500 repetitions of training per Tansig transfer function (Tansant Sigmoid) and Traingdm training algorithm (reduction gradient with momentum), is the best MLP network. In addition, the values of NSE, RMSE and MAE statistical indices for the network training stage are equal to 0.72, 0.22 and 0.15 respectively. Conclusion: Air pollution, the primary environmental challenge in Ahvaz, arises from the intersection of traffic and the oil industry. Its impacts on health and the environment necessitate comprehensive investigation. In this study, an MLP network was employed to predict CO gas concentration values in the air of Ahvaz city. The findings demonstrate that the network's accuracy and performance in forecasting CO gas concentration are at an optimal level. As this research progresses, it is recommended to extend the prediction to other gaseous pollutants and to employ optimization algorithms for determining the optimal structure of the artificial neural network Manuscript profile