Air Quality Assessment and Pollution Forecasting using Recurrent Artificial Neural Networks in Metropolitan Lima-Peru

The prediction of air pollution is of great importance in highly populated areas because it has a direct impact on both the management of the city's economic activity and the health of its inhabitants. In this work, the spatio-temporal behavior of air quality in Metropolitan Lima was evaluated and predicted using the recurrent artificial neural network known as Long-Short Term Memory networks (LSTM). The LSTM was implemented for the hourly prediction of PM10 based on the past values of this pollutant and three meteorological variables obtained from five monitoring stations. The model was evaluated under two validation schemes: the hold-out (HO) and the blocked-nested cross-validation (BNCV). The simulation results show that periods of low PM10 concentration are predicted with high precision. Whereas, for periods of high contamination, the LSTM network with BNCV has better predictability performance. In conclusion, recurrent artificial neural networks with BNCV adapt more precisely to critical pollution episodes and have better performance to forecast this type of environmental data, and can also be extrapolated to other pollutants.

The structure dilemma in biological and artificial neural networks

Brain research up to date has revealed that structure and function are highly related. Thus, for example, studies have repeatedly shown that the brains of patients suffering from schizophrenia or other diseases have a different connectome compared to healthy people. Apart from stochastic processes, however, an inherent logic describing how neurons connect to each other has not yet been identified. We revisited this structural dilemma by comparing and analyzing artificial and biological-based neural networks. Namely, we used feed-forward and recurrent artificial neural networks as well as networks based on the structure of the micro-connectome of C. elegans and of the human macro-connectome. We trained these diverse networks, which markedly differ in their architecture, initialization and pruning technique, and we found remarkable parallels between biological-based and artificial neural networks, as we were additionally able to show that the dilemma is also present in artificial neural networks. Our findings show that structure contains all the information, but that this structure is not exclusive. Indeed, the same structure was able to solve completely different problems with only minimal adjustments. We particularly put interest on the influence of weights and the neuron offset value, as they show a different adaption behaviour. Our findings open up new questions in the fields of artificial and biological information processing research.

New method for wind potential prediction using recurrent artificial neural networks

The aim of the study is to find the right architecture of the NARX neural network, in order to perform the daily prediction of the maximum wind speed of Laayoune city. We relied on the Levenberg-Marquardt optimization algorithm. The RMSE error metric showed that NARX-SP outperforms NARX-P.

Prediction of Brazilian Electric Energy Price Using Recurrent Artificial Neural Networks and Correction Filter

O Preço de Liquidação das Diferenças (PLD) é uma variável utilizada para determinar o valor a ser cobrado pelos volumes de energia que serão liquidados na Câmara de Comercialização de Energia Elétrica (CCEE), e é atualizado semanalmente. Seu cálculo é baseado em modelos estatísticos e matemáticos de otimização, e, portanto, apresenta um comportamento altamente não-linear. Este trabalho propõe, por meio de uma arquitetura recorrente de redes neurais artificiais LSTM e um filtro corretivo, a predição do preço do PLD uma semana à frente, buscando obter as melhores variáveis de entrada, a fim de contornar problemas recorrentes que aparecerem com o uso de redes recursivas em séries temporais. O resultado mostra como a obtenção das variáveis corretas acarretam em uma predição confiável do PLD.

An Amazon stingless bee foraging activity predicted using recurrent artificial neural networks and attribute selection

Bees play a key role in pollination of crops and in diverse ecosystems. There have been multiple reports in recent years illustrating bee population declines worldwide. The search for more accurate forecast models can aid both in the understanding of the regular behavior and the adverse situations that may occur with the bees. It also may lead to better management and utilization of bees as pollinators. We address an investigation with Recurrent Neural Networks in the task of forecasting bees’ level of activity taking into account previous values of level of activity and environmental data such as temperature, solar irradiance and barometric pressure. We also show how different input time windows, algorithms of attribute selection and correlation analysis can help improve the accuracy of our model.