Nonlinear Artificial Neural Network Research Articles

Modeling Antibacterial Activity with Machine Learning and Fusion of Chemical Structure Information with Microorganism Metabolic Networks.

Predicting the activity of new chemical compounds over pathogenic microorganisms with different metabolic reaction networks (MRN s) is an important goal due to the different susceptibility to antibiotics. The ChEMBL database contains >160 000 outcomes of preclinical assays of antimicrobial activity for 55 931 compounds with >365 parameters of activity (MIC, IC50, etc.) and >90 bacteria strains of >25 bacterial species. In addition, the Leong and Barabàsi data set includes >40 MRNs of microorganisms. However, there are no models able to predict antibacterial activity for multiple assays considering both drug and MRN structures at the same time. In this work, we combined perturbation theory, machine learning, and information fusion techniques to develop the first PTMLIF model. The best linear model found presented values of specificity = 90.31/90.40 and sensitivity = 88.14/88.07 in training/validation series. We carried out a comparison to nonlinear artificial neural network (ANN) techniques and previous models from the literature. Next, we illustrated the practical use of the model with an experimental case of study. We reported for the first time the isolation and characterization of terpenes from the plant Cissus incisa. The antibacterial activity of the terpenes was experimentally determined. The more active compounds were phytol and α-amyrin, with MIC = 100 μg/mL for Vancomycin-resistant Enterococcus faecium and Acinetobacter baumannii resistant to carbapenems. These compounds are already known from other sources. However, they have been isolated and evaluated for the first time here against several strains of multidrug-resistant bacteria including World Health Organization (WHO) priority pathogens. Last, we used the model to predict the activity of these compounds versus other microorganisms with different MRNs in order to find other potential targets.

Determining China's CO2 emissions peak with a dynamic nonlinear artificial neural network approach and scenario analysis

The global community and the academic world have paid great attention to whether and when China's carbon dioxide (CO2) emissions will peak. Our study investigates the issue with the Nonlinear Auto Regressive model with exogenous inputs (NARX), a dynamic nonlinear artificial neural network that has not been applied previously to this question. The key advance over previous models is the inclusion of feedback mechanisms such as the influence of past CO2 emissions on current emissions. The results forecast that the peak of China's CO2 emissions will occur in 2029, 2031 or 2035 at the level of 10.08, 10.78 and 11.63 billion tonnes under low-growth, benchmark moderate-growth, and high-growth scenarios. Based on the methodology of the mean impact value (MIV), we differentiate and rank the importance of the influence factors on CO2 emissions whereas previous studies included but did not rank factors. We suggest that China should choose the moderate growth development road and achieve its peak target in 2031, focusing on reducing CO2 emissions as a percent of GDP, less carbon-intensive industrialization, and choosing technologies that reduce CO2 emissions from coal or increasing the use of less carbon-intensive fuels.

DeepSpectra: An end-to-end deep learning approach for quantitative spectral analysis

Learning patterns from spectra is critical for the development of chemometric analysis of spectroscopic data. Conventional two-stage calibration approaches consist of data preprocessing and modeling analysis. Misuse of preprocessing may introduce artifacts or remove useful patterns and result in worse model performance. An end-to-end deep learning approach incorporated Inception module, named DeepSpectra, is presented to learn patterns from raw data to improve the model performance. DeepSpectra model is compared to three CNN models on the raw data, and 16 preprocessing approaches are included to evaluate the preprocessing impact by testing four open accessed visible and near infrared spectroscopic datasets (corn, tablets, wheat, and soil). DeepSpectra model outperforms the other three convolutional neural network models on four datasets and obtains better results on raw data than in preprocessed data for most scenarios. The model is compared with linear partial least square (PLS) and nonlinear artificial neural network (ANN) methods and support vector machine (SVR) on raw and preprocessed data. The results show that DeepSpectra approach provides improved results than conventional linear and nonlinear calibration approaches in most scenarios. The increased training samples can improve the model repeatability and accuracy.

Short and Long-Term Time Series Forecasting Stochastic Analysis for Slow Dynamic Processes

This paper intends to develop suitable methods to provide likely scenarios in order to support decision making for slow dynamic processes such as the underlying of agribusiness. A new method to analyze the short- and long-term time series forecast and to model the behavior of the underlying process using nonlinear artificial neural networks (ANN) is presented. The algorithm can effectively forecast the time-series data by stochastic analysis (Monte Carlo) of its future behavior using fractional Gaussian noise (fGn). The algorithm was used to forecast country risk time series for several countries, both for short term that is 30 days ahead and long term 350 days ahead scenarios.

Nano Zero-Valent Aluminum (nZVAl) Preparation, Characterization, and Application for the Removal of Soluble Organic Matter with Artificial Intelligence, Isotherm Study, and Kinetic Analysis

Zero-valent metals proved high reactivity to adsorb and degrade various contaminants removal. The chemically prepared nZVAl was characterized using UV-Vis spectrum, X-ray diffraction (XRD), and scanning electron microscope (SEM). This investigation explores the adsorption effect of nZVAl powder toward soluble organic compounds exemplified by chemical oxygen demand (COD) standard solution. The effect of different operating parameters was studied to identify the best removal conditions. All variable and covariable data were introduced to build statistical models. The effect of the operating parameter was studied at different pH (3-10), nZVAl dosages (0.1-0.8 g), at different times (5-120 minutes), stirring rate (50-400 RPM), and initial COD concentration (100-800 mg/L). The obtained results displayed that nZVAl is effective in the removal of standard COD solutions, where the removal percentages were 56% and 96% for 800 ± 18.0 and 100 ± 11.8 mg/L COD, respectively, at 10 minutes after using nZVAl dry dosage 0.6 g/L, pH 8, and rate 100 rpm. Also, the effect of nZVAl on other wastewater contaminants removal was studied and compared with Egyptian law for draining wastewater into nonfresh water (drainage-lakes-ponds) No. 48 of 1982 limits. The results of adsorption isotherm and kinetic model of COD fitted well to Freundlich isotherm and pseudo second order, respectively. Nonlinear artificial intelligence neural network (ANN) importance data agree with linear response surface methodologies (RSM) in simulating the adsorption of COD onto nZVAl indicating that the most significant coverable is adsorbent dose. Finally, this study appropriates using nZVAl in highly contaminated wastewater rather than other chemical and biological processes.

Economically Optimal Control of a Cold Room Using an Artificial Neural Network and Dynamic Programming

We consider the economically optimal control of a cold store with a single cold room. The thermal inertia of a cold room acts as an energy storage and can therefore be used for economic optimization in the presence of a dynamic electricity price, under a bounding constraint on the internal temperature of the cold room. However, a high number of frost production startups may induce premature wear of the cold store’s compressors. Since the thermal losses are a function of the internal temperature of the cold room, conventional inventory management solving techniques are not suited for this problem. In this paper, we use an artificial neural network as temperature forecast. A dynamic programming algorithm is used to solve the model that includes the non-linear artificial neural network temperature forecast and a fixed cost at each compressor startup. This allows us to solve industrial instances of the problem optimally and within reasonable time. We show the interest of solving the problem optimally as opposed to using a conventional hysteresis-based control method, and discuss the opportunity of using an dynamic hourly price based on the electricity market instead of a traditional contracted price.

Determination of sulfur dioxide content in osmotically dehydrated papaya and its classification by near infrared spectroscopy

Sulfur dioxide (SO2) is used as a preservative in osmotically dehydrated papaya to improve product quality and extend shelf-life. The potential of near infrared spectroscopy, as a rapid method, was investigated to determine sulfur dioxide in osmotically dehydrated papaya. Commercial and laboratory osmotically dehydrated papaya samples were selected to determine the sulfur dioxide content using the Monier–Williams method. From the total of 350 samples, subsets were selected randomly for the calibration set (n=250) and validation set (n = 100). Near infrared spectra in the region 800–2400 nm were measured on the samples of osmotically dehydrated papaya. Quantitative analyses of sulfur dioxide in the osmotically dehydrated papaya and their qualitative analyses were carried out using multivariate analysis. Before developing models, a second derivative spectral pretreatment was applied to the original spectral data. Subsequently, two wavelength interval selection methods, namely moving window partial least squares regression (MWPLSR) and searching combination moving window partial least squares (SCMWPLS), were applied to determine the suitable input wavelength variables. For quantitative analysis, three linear models (partial least squares regression, MWPLSR and SCMWPLS) and a non-linear artificial neural network model were applied to develop predictive models. The results showed that the artificial neural network model produced the best performance, with correlation coefficient (R) and root mean square error of prediction values of 0.937 and 114.53 mg SO2 kg−1, respectively. Qualitative models were developed using partial least squares-discriminant analysis and soft independent modeling of class analogy (SIMCA) for the optimized combination of informative regions of the near infrared spectra to classify osmotically dehydrated papaya into three groups based on sulfur dioxide. The SIMCA in combination with SCMWPLS model had the highest correct classification rate (96%). The study demonstrated that near infrared spectroscopy combined with SCMWPLS is a powerful procedure for both quantitative and qualitative analyses of osmotically dehydrated papaya. Therefore, it was demonstrated that near infrared spectroscopy could be effective tools for food quality and safety evaluation in food industry.

The research about micromachining accuracy control based on fuzzy evaluation algorithm

SummaryThe calculation method of nonlinear artificial neural network based on fuzzy evaluation algorithm is introduced. This method is used to evaluate the precision of micro machined machining and to find the influence factors of the precision of micromachining. The traditional machining precision control mainly adopts the linear control strategy, but the micromachining involves the mechanical characteristics of the micro field, and many linear machine adding precision control methods cannot effectively control the precision of the micromachining. Using fuzzy evaluation method, large data cluster training method can be used to fit the control curve of micromachining accuracy. The nonlinear mapping of the fuzzy evaluation algorithm is realized by the large sample training method, and the objective and effective parameter evaluation and precision control of the machining precision in the micro machined field are realized. It is proved that the fuzzy evaluation algorithm can accurately evaluate the precision of micro machined machining. Learning through large data samples can effectively improve the topology practicability and rationality of the algorithm.

A nonlinear ANN equalizer with mini-batch gradient descent in 40Gbaud PAM-8 IM/DD system

Abstract We propose and validate a novel nonlinear artificial neural network (ANN) equalizer for PAM-8 transmission in IM/DD system. Mini-batch gradient descent is introduced to efficiently train ANN equalizer. Using the proposed ANN equalizer, we successfully transmit a 40Gbaud PAM-8 signal over 4-km SMF with BER under the threshold of 3.8 × 10−3 and over 10-km SMF with BER under the threshold of 1 × 10−2. We also elaborately compare the proposed ANN equalizer with other methods including LMS equalizer, Volterra equalizer and look-up table (LUT). Experimental results indicate that ANN achieves the best performance that is slightly superior to Volterra equalizer with computational complexity half reduced. To the best of our knowledge, this is the first time to adopt mini-batch gradient descent to train ANN equalizer in IM/DD system.

STREAMFLOW FORECASTING IN BUKIT MERAH WATERSHED BY USING ARIMA AND ANN

This paper presents the application of linear and non-linear time series modeling approaches for simulating and forecasting streamflow at three stations located in three different rivers namely Kurau River, Ara River and Krian River of Bukit Merah watershed of Malaysia. The performance of linear autoregressive integrated moving average (ARIMA) model and non-linear artificial neural networks (ANN) model in forecasting the monthly streamflow of Malaysian river basins has been evaluated based on mean absolute percentage error (MAPE), root mean squared error (RMSE) and coefficient of determination (R2). The results show that both ARIMA and ANN methods are suitable for streamflow forecasting. However, ANN is better than ARIMA in dealing with short-memory streamflow data. In addition, ANN method is more flexible to use against the inconsistent data.Keywords: time series, streamflow forecasting, ARIMA, ANN, Bukit Merah

PTML Combinatorial Model of ChEMBL Compounds Assays for Multiple Types of Cancer.

Determining the target proteins of new anticancer compounds is a very important task in Medicinal Chemistry. In this sense, chemists carry out preclinical assays with a high number of combinations of experimental conditions (c j). In fact, ChEMBL database contains outcomes of 65 534 different anticancer activity preclinical assays for 35 565 different chemical compounds (1.84 assays per compound). These assays cover different combinations of c j formed from >70 different biological activity parameters ( c0), >300 different drug targets ( c1), >230 cell lines ( c2), and 5 organisms of assay ( c3) or organisms of the target ( c4). It include a total of 45 833 assays in leukemia, 6227 assays in breast cancer, 2499 assays in ovarian cancer, 3499 in colon cancer, 3159 in lung cancer, 2750 in prostate cancer, 601 in melanoma, etc. This is a very complex data set with multiple Big Data features. This data is hard to be rationalized by researchers to extract useful relationships and predict new compounds. In this context, we propose to combine perturbation theory (PT) ideas and machine learning (ML) modeling to solve this combinatorial-like problem. In this work, we report a PTML (PT + ML) model for ChEMBL data set of preclinical assays of anticancer compounds. This is a simple linear model with only three variables. The model presented values of area under receiver operating curve = AUROC = 0.872, specificity = Sp(%) = 90.2, sensitivity = Sn(%) = 70.6, and overall accuracy = Ac(%) = 87.7 in training series. The model also have Sp(%) = 90.1, Sn(%) = 71.4, and Ac(%) = 87.8 in external validation series. The model use PT operators based on multicondition moving averages to capture all the complexity of the data set. We also compared the model with nonlinear artificial neural network (ANN) models obtaining similar results. This confirms the hypothesis of a linear relationship between the PT operators and the classification as anticancer compounds in different combinations of assay conditions. Last, we compared the model with other PTML models reported in the literature concluding that this is the only one PTML model able to predict activity against multiple types of cancer. This model is a simple but versatile tool for the prediction of the targets of anticancer compounds taking into consideration multiple combinations of experimental conditions in preclinical assays.

Artificial intelligence driven process optimization for cleaner production of biomass with co-valorization of wastewater and flue gas in an algal biorefinery

In the present study, an optimized process involving integrated events of flue gas CO2 sequestration and wastewater utilization was developed for improved microalgal biomass production in a cleaner fashion. An artificial neural network (ANN) combined with genetic algorithm (GA) optimization tool was employed for predicting optimal process conditions for enhancing the biomass of the green microalga, Scenedesmus sp., using domestic wastewater as culture medium and coal-fired flue gas as carbon source in an integrated process chain. A 24 central composite design involving four independent process parameters such as light intensity, photoperiod, temperature and initial pH as input variables with biomass productivity as the output was used to construct a non-linear ANN model followed by GA tool to predict optimal combinations of process conditions. Among the tested neural network architecture, 4-12-1 ANN topology was found to be the optimal network architecture in terms of maximum correlation coefficient (R = 0.9947) and minimum mean square error as the performance indexes. On employing the optimized ANN model as fitness function in GA tool, the optimum values of the process parameters for efficient biomass production were as follows: light intensity = 124 μmol m−2 s−1, photoperiod; L:D = 17:7 (h), temperature = 27.5 °C and initial pH = 9.5. These parameters improved the algal biomass productivity by about 57%, CO2 sequestration rate of 578.1 ± 23.1 mg L−1 d−1 and chemical oxygen demand (COD) reduction of 95.9 ± 2.4% were achieved. The optimized process yielded biomass with lipid content and productivity of 34.6% and 106.4 mg L−1 d−1 respectively. The biofuel assessment from the fatty acid methyl ester profile obtained also conformed to the international standard specifications for biodiesel.

Modelling asphalt pavement analyzer rut depth using different statistical techniques

Different computer-operated temperature-controlled equipments are used to predict asphalt mixture performance in laboratory at approximately similar conditions to actual in-service pavements. Asphalt pavement analyzer (APA) is one such multi-functional sophisticated equipment. Wearing course layers are primarily attributable to be deteriorated by rutting distress in flexible asphalt pavements. Therefore, 30 representative wearing course mixtures commonly used in road pavements were tested for rutting distress using APA. Mixtures were evaluated based on their rutting performance and found that consistent ranking has been observed at temperature conditions of 40°C and 50°C. It was observed that APA rut depth could be effectively modelled using statistical techniques of multiple non-linear regression and artificial neural network. Novel aggregate material indices designated as ‘aggregate source index’ and ‘aggregate gradation index’ were developed in this study. Based on past literature; temperature, aggregate source index (ASI), aggregate gradation index, bitumen penetration values, and number of APA loading cycles/strokes were included as independent variables in the model developed for selected asphalt mixtures. It has been concluded from the values of non-linear regression parameters that temperature has the highest significance for rut depth prediction, followed by the bitumen penetration value, ASI, aggregate gradation index and number of loading cycles. Statistical parameters i.e. coefficient of determination, root mean squared error, values account for, and mean absolute percentage error showed that the artificial neural network technique better predicts the APA rut depth as compared to multiple non-linear regression techniques.

Combination Forecasting of Energy Demand in the UK

In more deregulated markets such as the UK, demand forecasting is vital for the electric industry as it is used to set electricity generation and purchasing, establishing electricity prices, load switching and demand response. In this paper we produce improved short-term forecasts of the demand for energy produced from five different sources in the UK averaging from a set of 6 univariate and multivariate models. The forecasts are averaged using six different weighting functions including Simple Model Averaging (SMA), Granger-Ramanathan Model Averaging (GRMA), Bayesian Model Averaging (BMA), Smoothing Akaike (SAIC), Mallows Weights (MMA) and Jackknife (JMA). Our results show that model averaging gives always a lower Mean Square Forecast Error (MSFE) than the best/optimal models within each class however selected. For example, for Coal, Wind and Hydro generated Electricity forecasts generated with model averaging, we report a MSFE about 12% lower than that obtained using the best selected individual models. Among these, the best individual forecasting models are the Non-Linear Artificial Neural Networks and the Vector Autoregression and that models selected by the Jackknife have often superior performance. However, MMA averaged forecasts almost always beat the predictions obtained from any of the individual models however selected, and those generated by other model averaging techniques.

Stochastic Forecasting Models of the Monthly Streamflow for the Blue Nile at Eldiem Station

Egypt is almost totally dependent on the River Nile for satisfying about 95% of its water requirements. The River Nile has three main tributaries: White Nile, Blue Nile, and River Atbara. The Blue Nile contributes about 60% of total annual flow reached the River Nile at Aswan High Dam. The goal of this research is to develop a reliable stochastic model for the monthly streamflow of the Blue Nile at Eldiem station, where the Grand Ethiopian Renaissance Dam (GERD) is currently under construction with a storage capacity of about 74 billion m3. The developed model may help to carry out a reliable study on the filling scenarios of GERD reservoir and to minimize its expected negative side effects on Sudan and Egypt. The linear models: Deseasonalized AutoRegressive Moving Average (DARMA) model, Periodic AutoRegressive Moving Average (PARMA) model and Seasonal AutoRegressive Integrated Moving Average (SARIMA) model; and the nonlinear Artificial Neural Network (ANN) model are selected for modeling monthly streamflow at Eldiem station. The performance of various models during calibration and validation were evaluated using the statistical indices: Mean Absolute Error, Root Mean Square Error and coefficient of determination (R2) which indicate the strength of fitting between observed and forecasted values. The results show that the performance of the nonlinear model (ANN) was much better than all investigated linear models (DARMA, PARMA and SARIMA) in forecasting the monthly flow discharges at Eldiem station.

Day-ahead forecasting of grid carbon intensity in support of heating, ventilation and air-conditioning plant demand response decision-making to reduce carbon emissions

Electrical heating, ventilation and air-conditioning loads in buildings are suitable candidates for use in demand response activity. This paper demonstrates a method to support planned demand response actions intended explicitly to reduce carbon emissions. Demand response is conventionally adopted to aid the operation of electricity grids and can lead to greater efficiency; here it is planned to target times of day when electricity is generated with high carbon intensity. Operators of heating, ventilation and air-conditioning plant and occupants of conditioned spaces can plan when to arrange shutdown of plant once they can foresee the opportune time of day for carbon saving. It is shown that the carbon intensity of the mainland UK electricity grid varies markedly throughout the day, but that this tends to follow daily and weekly seasonal patterns. To enable planning of demand response, 24 h ahead forecast models of grid carbon intensity are developed that are not dependent on collecting multiple exogenous data sets. In forecasting half-hour periods of high carbon intensity either linear autoregressive or non-linear artificial neural network models can be used, but a daily seasonal autoregressive model is shown to provide a 20% improvement in carbon reduction. Practical application: The forecast method demonstrated in the paper would enable building operators to plan demand response activity to target times of high carbon intensity on the UK electricity grid. The method would be easy to implement as the only data required are publicly available.

Photocatalytic ozonation of ciprofloxacin from aqueous solution using TiO 2 /MMT nanocomposite: Nonlinear modeling and optimization of the process via artificial neural network integrated genetic algorithm

Abstract The degradation of ciprofloxacin (CIP) via photocatalytic ozonation process was investigated. TiO2 nanoparticles as an impressive photocatalyst were immobilized on the surface of montmorillonite (MMT). The main aim of the present work was to model and develop a relationship among the CIP degradation efficiency DE (%) and the key parameters affecting the complex photocatalytic ozonation process. The second object was to investigate the mutual effects of operating parameters on CIP DE (%) and express and optimum conditions that maximize CIP DE (%). To achieve these goals a nonlinear modeling and optimization of the process through artificial neural network (ANN) integrated genetic algorithm (GA) was successfully developed. The mutual effects of the input parameters including initial pH, ozone flow rate, initial CIP concentration, catalyst dosage and reaction time on CIP DE (%) were studied using the surface plots predicted with the nonlinear ANN model regarding the synergistic mechanism. GA was used as a nonlinear optimization method to determine the optimum conditions of input parameters. The good agreement between the experimental and GA predicted DE (%) under the optimum conditions confirmed that ANN integrated GA system is able to successfully model and optimize the photocatalytic ozonation performance.

DendroTools: R package for studying linear and nonlinear responses between tree-rings and daily environmental data

We introduce in this paper the dendroTools R package for studying the statistical relationships between tree-ring parameters and daily environmental data. The core function of the package is daily_response(), which works by sliding a moving window through daily environmental data and calculating statistical metrics with one or more tree ring proxies. Possible metrics are correlation coefficient, coefficient of determination and adjusted coefficient of determination. In addition to linear regression, it is possible to use a nonlinear artificial neural network with the Bayesian regularization training algorithm (brnn). dendroTools provides the opportunity to use daily climate data and robust nonlinear functions for the analysis of climate-growth relationships. Models should thus be better adapted to the real (continuous) growth of trees and should gain in predictive capabilities. The dendroTools R package is freely available in the CRAN repository. The functionality of the package is demonstrated on two examples, one using a mean vessel area (MVA) chronology and one a traditional tree-ring width (TRW).

Preparation of Drug Eluting Natural Composite Scaffold Using Response Surface Methodology and Artificial Neural Network Approach.

Silk fibroin/xanthan composite was investigated as a suitable biomedical material for controlled drug delivery, and blending ratios of silk fibroin and xanthan were optimized by response surface methodology (RSM) and artificial neural network (ANN) approach. A non-linear ANN model was developed to predict the effect of blending ratios, percentage swelling and porosity of composite material on cumulative percentage release. The efficiency of RSM was assessed against ANN and it was found that ANN is better in optimizing and modeling studies for the fabrication of the composite material. In-vitro release studies of the loaded drug chloramphenicol showed that the optimum composite scaffold was able to minimize burst release of drug and was followed by controlled release for 5days. Mechanistic study of release revealed that the drug release process is diffusion controlled. Moreover, during tissue engineering application, investigation of release pattern of incorporated bioactive agent is beneficial to predict, control and monitor cellular response of growing tissues. This work also presented a novel insight into usage of various drug release model to predict material properties. Based on the goodness of fit of the model, Korsmeyer-Peppas was found to agree well with experimental drug release profile, which indicated that the fabricated material has swellable nature. The chloramphenicol (CHL) loaded scaffold showed better efficacy against gram positive and gram negative bacteria. CHL loaded SFX55 (50:50) scaffold shows promising biocomposite for drug delivery and tissue engineering applications.

Assessing the impact of PM2.5 on respiratory disease using artificial neural networks.

Understanding the impact on human health during peak episodes in air pollution is invaluable for policymakers. Particles less than PM2.5 can penetrate the respiratory system, causing cardiopulmonary and other systemic diseases. Statistical regression models are usually used to assess air pollution impacts on human health. However, when there are databases missing, linear statistical regression may not process well and alternative data processing should be considered. Nonlinear Artificial Neural Networks (ANN) are not employed to research environmental health pollution even though another advantage in using ANN is that the output data can be expressed as the number of hospital admissions. This research applied ANN to assess the impact of air pollution on human health. Three well-known ANN were tested: Multilayer Perceptron (MLP), Extreme Learning Machines (ELM) and Echo State Networks (ESN), to assess the influence of PM2.5, temperature, and relative humidity on hospital admissions due to respiratory diseases. Daily PM2.5 levels were monitored, and hospital admissions for respiratory illness were obtained, from the Brazilian hospital information system for all ages during two sampling campaigns (2008-2011 and 2014-2015) in Curitiba, Brazil. During these periods, the daily number of hospital admissions ranged from 2 to 55, PM2.5 concentrations varied from 0.98 to 54.2 μg m-3, temperature ranged from 8 to 26 °C, and relative humidity ranged from 45 to 100%. Of the ANN used in this study, MLP gave the best results showing a significant influence of PM2.5, temperature and humidity on hospital attendance after one day of exposure. The Anova Friedman's test showed statistical difference between the appliance of each ANN model (p < .001) for 1 lag day between PM2.5 exposure and hospital admission. ANN could be a more sensitive method than statistical regression models for assessing the effects of air pollution on respiratory health, and especially useful when there is limited data available.