Search

This article proposes a method of estimating the concentrations of particulate matter (PM2.5 and PM10) using public data, including road-traffic closed-circuit television (CCTV) images, Smart Seoul City data sensor environment information, and Korea Meteorological Administration data. The region-of-interest images and full scenes derived from CCTV footage were used as the basis for the deep learning model, which combines a convolutional neural network and long short-term memory, to establish the particulate matter (PM) concentration prediction methodology. In the experiment, the prediction accuracies corresponding to various types of mean values were calculated by training the model with various mean measurement values for the surface PM2.5 and PM10 concentrations, as well as the corresponding CCTV images and weather data at different time points. In the experiments performed under relatively stable PM concentrations, R2 generally exceeded 0.9 and tended to increase with an increasing range of mean concentration values. In particular, in sections with rapid changes in the PM concentration within an hourly interval, higher R2 values were obtained by the model trained with the average PM concentrations of the time series before and after image capture, outperforming the method that used prior mean observation values and better reflecting the current PM concentration.

Heavy metals are very toxic and hazardous for human health. Onsite screening of heavy metal contaminated samples along with location-based automation data collection is a tedious job. Traditionally high-end equipment’s such as gas chromatography mass spectrometer (GC–MS) and atomic absorption spectrometers have been used to measure the concentration of different heavy metals in water samples but most of them are costly, bulky, and time consuming, and requires expert human intervention. This manuscript reports an ultra-portable, rapid, cost-effective, and easy-to-use solution for onsite heavy metal concentration measurement in drinking water samples. Presented solution combines off-the-shelf available chemical kits for heavy metal detection and developed spectrometer-based readout for concentration prediction, quality judgment, and automatic data collection. Two chemical kits for copper and iron detection have been imported form Merck and have been used for overall training and testing. The developed spectrometer has capability to work with smartphone-based android app and also can work in standalone mode. The developed spectrometer uses white light-emitting diode as a source and commercially imported spectral sensor (AS7262) for visible radiation reception. A low-power sub-GHZ-based wireless embedded platform has been developed and interfaced with source and detector. A power management module also has been designed to monitor the battery status and also to generate low battery indication. Overall modules has been packaged in custom designed enclosure to avoid external light interference. The developed system has been trained using standard buffer samples with known heavy metal concentrations and further tested for water samples collected from institute colony and nearby villages. The obtained results have been validated with commercially imported system from HANNA instruments, and it has been observed that developed system has shown excellent accuracy to predict heavy metal concentration (tested for Fe and Cu) in water samples.

The time series of blood glucose concentration in diabetic patients are time-varying, nonlinear, and non-stationary. In order to improve the accuracy of blood glucose prediction, a multi-scale combination short-term blood glucose prediction model was constructed by combining the variational mode decomposition (VMD) method, the kernel extreme learning machine (KELM), and the AdaBoost algorithm (VMD-ELM-AdaBoost). Firstly, the blood glucose concentration series were decomposed into a set of intrinsic modal functions (IMFs) with different scales by the VMD method. On this basis, the KELM neural network and AdaBoost algorithm are combined to predict each IMF component. Finally, the cumulative blood glucose concentration prediction value is obtained by accumulating the KELM-AdaBoost prediction results of each IMF. The time series of measured blood glucose concentration were used for experimental analysis; the experimental results show that the proposed VMD-KELM-AdaBoost method has higher prediction accuracy compared with the classical prediction models such as ELM, KELM, SVM, and LSTM. The proposed VMD-KELM-AdaBoost model can still achieve high prediction accuracy 60 min in advance (the mean values of RMSE, MAPE, and CC are about 10.1422, 4.8629%, and 0.8737 respectively); in Clarke error mesh analysis, the proportion of falling into A region is about 95.7%; the sensitivity and false alarm rate of early alarm of hypoglycemia were 94.8% and 7.7%, respectively. Graphical abstract We have proposed a new prediction model. In the first part, for reducing thenon-stationarity, the data of blood glucose concentration was decomposed as a series ofIMF by VMD. In the second part, a prediction model based KELM and Adaboost wasestablished. In the third part, the KELM-Adaboost model was used to predict each IMF,and the predicted values of all IMFS were superimposed to obtain the final predictionresult of blood glucose concentration.

The aim of this study is to show the robustness of a commercial direct compression system running continuously under relevant process conditions for over 120 h. The study will apply a representative commercial control strategy including real time monitoring of process parameters, rejection of OOS material, use of in-line NIR prediction for blend potency monitoring, application of mathematical RTD modelling for real-time prediction of product potency, and tablet concentration prediction by tablet NIR. The study showed that the process can produce in specification product robustly over significant operating times. Across 122 h of actual run time (including pauses) the process manufactured over 15 million tablets to give a product yield of 99.2%. Whilst the purpose of this study was to evaluate the process for over 120 h, this process is not limited to running longer periods of time given the performance of the equipment and the control strategy.

Air pollution poses a great threat to human health, and people are paying more and more attention to the prediction of air pollution. Prediction of air pollution helps people plan for their outdoor activities and helps protect human health. In this article, long-short term memory recurrent neural networks were used to predict the future concentration of air pollutants in Macau. In addition, meteorological data and data on the concentration of air pollutants were used. Moreover, in Macau, some air quality monitoring stations have less observed data, and some AQMSs less observed data of certain types of air pollutants. Therefore, the transfer learning and pre-trained neural networks were used to assist AQMSs with less observed data to generate neural network with high prediction accuracy. In this thesis, in most cases, LSTM RNNs initialized with transfer learning methods have higher prediction accuracy, used less training time than randomly initialized recurrent neural networks.

People have been suffering from air pollution for a decade in China, especially from PM2.5 (particulate matter with a diameter of less than 2.5 μm). Accurate prediction of air quality has great practical significance. In this paper, we propose a data-driven model, called as long short-term memory - fully connected (LSTM-FC) neural network, to predict PM2.5 contamination of a specific air quality monitoring station over 48 h using historical air quality data, meteorological data, weather forecast data, and the day of the week. Our predictive model consists of two components: (1) Using a long short-term memory (LSTM)-based temporal simulator to model the local variation of PM2.5 contamination and (2) Using a neural network-based spatial combinatory to capture spatial dependencies between the PM2.5 contamination of central station and that of neighbor stations. We evaluate our model on a dataset containing records of 36 air quality monitoring stations in Beijing from 2014/05/01 to 2015/04/30 and compare it with artificial neural network (ANN) and long short-term memory (LSTM) models on the same dataset. The results show that our LSTM-FC neural network model gives a better predictive performance.

Many countries are concerned about high particulate matter (PM) concentrations caused by rapid industrial development, which can harm both human health and the environment. To manage PM, the prediction of PM concentrations based on historical data is actively being conducted. Existing technologies for predicting PM mostly assess the model performance for the prediction of existing PM concentrations, however, PM must be forecast in advance, before it becomes highly concentrated and causes damage to the citizens living in the affected regions. Thus, it is necessary to conduct research on an index that can illustrate whether the PM concentration will increase or decrease. We developed a model that can predict whether the PM concentration might increase or decrease after a certain time, specifically for PM2.5 (fine PM) generated by anthropogenic volatile organic compounds. An algorithm that can select a model on an hourly basis, based on the long short-term memory (LSTM) and artificial neural network (ANN) models, was developed. The proposed algorithm exhibited a higher F1-score than the LSTM, ANN, or random forest models alone. The model developed in this study could be used to predict future regional PM concentration levels more effectively.

Achieving accurate predictions of urban NO2 concentration is essential for effectively control of air pollution. This paper selected the concentration of NO2 in Tianjin as the research object, concentrating predicting model based on Discrete Wavelet Transform and Long- and Short-Term Memory network (DWT-LSTM) for predicting daily average NO2 concentration. Five major atmospheric pollutants, key meteorological data, and historical data were selected as the input indexes, realizing the effective prediction of NO2 concentration in the next day. Firstly, the input data were decomposed by Discrete Wavelet Transform to increase the data dimension. Furthermore, the LSTM network model was used to learn the features of the decomposed data. Ultimately, Support Vector Regression (SVR), Gated Regression Unit (GRU), and single LSTM model were selected as comparison models, and each performance was evaluated by the Mean Absolute Percentage Error (MAPE). The results show that the DWT-LSTM model constructed in this paper can improve the accuracy and generalization ability of data mining by decomposing the input data into multiple components. Compared with the other three methods, the model structure is more suitable for predicting NO2 concentration in Tianjin.

At present, most coal-fired units use selective catalytic reduction of (SCR) and selective non-catalytic reduction of (SNCR) denitrification technology for NOx removal. Accurate prediction of NOx emission concentration of coal-fired units will not only help to further improve the regulation quality of denitrification control system, but also evaluate whether the data collected in the current site are true and accurate, and provide a basis for environmental protection departments to supervise and enforce the law of NOx emission from power plants. In this paper, based on the historical operation data of a 200MW circulating fluidized bed unit, by analyzing the factors affecting the NOx emission concentration, firstly, the correlation coefficient method is used to analyze the delay between each factor and the NOx emission concentration, then the BP neural network is used to model the two-stage intersection, the established NOx emission concentration prediction model is compared, and a more accurate NOx emission concentration prediction model is selected. Finally, the factors affecting NOx emission concentration are selected from a more accurate model. The results show that the root mean square error of the first kind of modeling is 0.023 less than that of the second kind of modeling, so the six input factors in the first kind of model are regarded as the best factors affecting the NOx emission concentration. The selected factors can be used to accurately predict the NOx emission concentration for a period of time in the future, which lays a foundation for more accurate control of SNCR denitrification control system.

The odor emitted from a wastewater treatment plant (WWTP) is an important environmental problem. An estimation of odor emission rate is difficult to detect and quantify. To address this, various approaches including the development of emission factors and measurement using a closed chamber have been employed. However, the evaluation of odor emission involves huge manpower, time, and cost. An artificial neural network (ANN) is recognized as an efficient method to find correlations between nonlinear data and prediction of future data based on these correlations. Due to its usefulness, ANN is used to solve complicated problems in various disciplines of sciences and engineering. In this study, a method to predict the odor concentration in a WWTP using ANN was developed. The odor concentration emitted from a WWTP was predicted by the ANN based on water quality data such as biological oxygen demand, dissolved oxygen, and pH. The water quality and odor concentration data from the WWTP were measured seasonally in spring, summer, and autumn and these were used as input variations to the ANN model. The odor predicted by the ANN model was compared with the measured data and the prediction accuracy was estimated. Suggestions for improving prediction accuracy are presented.

Air quality prediction is a global hot issue, and PM2.5 is an important factor affecting air quality. Due to complicated causes of formation, PM2.5 prediction is a thorny and challenging task. In this paper, a novel deep learning model named temperature-based deep belief networks (TDBN) is proposed to predict the daily concentrations of PM2.5 for the next day. Firstly, the location of PM2.5 concentration prediction is Chaoyang Park in Beijing of China from January 1, 2018 to October 27, 2018. The auxiliary variables are selected as input variables of TDBN by Partial Least Square (PLS), and the corresponding data is divided into three independent sections: training samples, validating samples and testing samples. Secondly, the TDBN is composed of temperature-based restricted Boltzmann machine (RBM), where temperature is considered as an effective physical parameter in energy balance of training RBM. The structural parameters of TDBN are determined by minimizing the error in the training process, including hidden layers number, hidden neurons and value of temperature. Finally, the testing samples are used to test the performance of the proposed TDBN on PM2.5 prediction, and the other similar models are tested by the same testing samples for convenience of comparison with TDBN. The experimental results demonstrate that TDBN performs better than its peers in root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R2).

Machine learning has been a powerful tool to establish models for timely prediction of PM2.5 concentration, which supports the policy makers to take control actions and facilitates sustainable development of cities by protection on human health. Information of adjacent sites can be a good indicator to represent the regional pattern of PM2.5 pollution, however it has not been widely investigated in most existing prediction models. In this study, a novel hybrid prediction model on PM2.5 concentration taking advantage of appropriate adjacent site observations was proposed. This hybrid model combined long short-term memory neural network (LSTM) and a specific convolutional neural network (CNN) with 1 * 1 kernel size to address the traditional challenge of time-series data and discrete data. The experimental results showed that the novel hybrid model presented the highest prediction accuracy and the lowest error (R2=0.92-0.94, RMSE=8.54-8.93μg/m3, MAE=5.76-6.52μg/m3) compared with single models (R2=0.81-0.92, RMSE=9.01-13.24μg/m3, MAE=6.79-9.34μg/m3). The hybrid model is also compared with other similar studies, and it demonstrates a significantly enhanced ability to capture the peak values of PM2.5 concentration by information from appropriate adjacent sites.

For hourly PM 2.5 concentration prediction, accurately capturing the data patterns of external factors that affect PM 2.5 concentration changes, and constructing a forecasting model is one of efficient means to improve forecasting accuracy. In this study, a novel hybrid forecasting model based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and deep temporal convolutional neural network (DeepTCN) is developed to predict PM 2.5 concentration, by modeling the data patterns of historical pollutant concentrations data, meteorological data, and discrete time variables’ data. Taking PM 2.5 concentration of Beijing as the sample, experimental results showed that the forecasting accuracy of the proposed CEEMDAN-DeepTCN model is verified to be the highest when compared with the statistics-based models, traditional machine learning models, the popular deep learning models and several existing hybrid models. The new model has improved the capability to model the PM 2.5 -related factor data patterns, and can be used as a promising tool for forecasting PM 2.5 concentrations.

Following the Process Analytical Technology (PAT) of the Food and Drug Administration (FDA), drug manufacturers are encouraged to develop innovative techniques in order to monitor and understand their processes in a better way. Within this framework, it has been demonstrated that Raman spectroscopy coupled with chemometric tools allow to predict critical parameters of mammalian cell cultures in-line and in real time. However, the development of robust and predictive regression models clearly requires many batches in order to take into account inter-batch variability and enhance models accuracy. Nevertheless, this heavy procedure has to be repeated for every new line of cell culture involving many resources. This is why we propose in this paper to develop global regression models taking into account different cell lines. Such models are finally transferred to any culture of the cells involved. This article first demonstrates the feasibility of developing regression models, not only for mammalian cell lines (CHO and HeLa cell cultures), but also for insect cell lines (Sf9 cell cultures). Then global regression models are generated, based on CHO cells, HeLa cells, and Sf9 cells. Finally, these models are evaluated considering a fourth cell line(HEK cells). In addition to suitable predictions of glucose and lactate concentration of HEK cell cultures, we expose that by adding a single HEK-cell culture to the calibration set, the predictive ability of the regression models are substantially increased. In this way, we demonstrate that using global models, it is not necessary to consider many cultures of a new cell line in order to obtain accurate models. Biotechnol. Bioeng. 2017;114: 2550-2559. © 2017 Wiley Periodicals, Inc.

PM2.5 concentration have received considerable attention from meteorologists, who are able to notify the public and take precautionary measures to prevent negative effects on health. Therefore, establishing an efficient early warning system plays a critical role in fostering public health in heavily polluted areas. In this study, ensemble empirical mode decomposition and least square support vector machine (EEMD-LSSVM) based on Phase space reconstruction (PSR) is proposed for day-ahead PM2.5 concentration prediction, according to the application of a decomposition-ensemble learning paradigm. The main methods of the proposed model mainly include: first, EEMD is presented to decompose the original data of PM2.5 concentration into some intrinsic model functions (IMFs); second, PSR is applied to determine the input form of each extracted component; third, LSSVM, an effective forecasting tool, is used to predict all reconstructed components independently; finally, another LSSVM is employed to aggregate all predicted components into ensemble results for the final prediction. The empirical results show that this proposed model can outperform the comparison models and can significantly improve the prediction performance in terms of higher predictive and directional accuracy.

To improve the predicting accuracy of PM10 concentration prediction, this paper presents a combined prediction model of PM10 concentration based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), sample entropy (SE) and long short-term memory (LSTM). The PM10 concentration time series is decomposed into a series of restructured subsequences with obvious complexity differences by CEEMDAN-SE firstly. Then, by adding meteorological parameters to each different restrict-ured subsequence, the LSTM prediction model is built. By adding the prediction results, the final results are got. Meanwhile, the data of four monitoring stations in Tangshan city is used to implement simulation experiment. Experiment results confirm that the proposed prediction model compares with other prediction models to show high prediction precision, and good universality.

Accurate PM2.5 concentration prediction is crucial for protecting public health and improving air quality. As a popular deep learning model, deep belief network (DBN) for PM2.5 concentration prediction has received increasing attention due to its effectiveness. However, the DBN structure parameters that have a significant impact on prediction accuracy and computation time are hard to be determined. To address this issue, a modified grey wolf optimization (MGWO) algorithm is proposed to optimize the DBN structure parameters containing number of hidden nodes, learning rate, and momentum coefficient. The methodology modifies the basic grey wolf optimization (GWO) algorithm using the nonlinear convergence and position update strategies, and then utilizes the training error of the DBN to calculate the fitness function of the MGWO algorithm. Through the multiple iterations, the optimal structure parameters are obtained, and a suitable predictor is finally generated. The proposed prediction model is validated on a real application case. Compared with the other prediction models, experimental results show that the proposed model has a simpler structure but higher prediction accuracy.

Parallel factor analysis is a widely used method to extract qualitative and quantitative information of the analyte of interest from fluorescence emission-excitation matrix containing unknown components. Big amplitude of scattering will influence the results of parallel factor analysis. Many methods of eliminating scattering have been proposed. Each of these methods has its advantages and disadvantages. The combination of symmetrical subtraction and interpolated values has been discussed. The combination refers to both the combination of results and the combination of methods. Nine methods were used for comparison. The results show the combination of results can make a better concentration prediction for all the components.

Effective monitoring and prediction of PM2.5 concentration is of great practical significance to People’s Daily life and social development. In order to facilitate accurate prediction of the average daily concentration of PM2.5, a Feature Selection Algorithm Based on Pseudo F Statistic (FSPF) was proposed in this paper, and a PM2.5 concentration prediction model was combined with support vector regression. When constructing the model, all features are not selected into the model, and the pseudo-F statistic featu re selection algorithm is used to obtain more important feature variables related to PM2.5 concentration. On this basis, support vector regression based on different kernel functions is used to predict PM2.5 concentration. Through real data analysis, the hyperbolic tangent kernel support vector regression model (FSPF-SVR) selected by pseudo-F statistics features has the highest prediction accuracy when predicting the average PM2.5 concentration of Beijing in November and December 2017, with mean square errors of 1.441⋅10−3mg2/m6 and 1.307⋅10−3mg2/m6 respectively, which is better than the all-variable prediction model. At the same time, it is also better than random forest, weighted K nearest neighbor, Bagging regression, artificial neural network and other current mainstream machine learning regression prediction algorithms.

In the era of big data, a variety of factors (particularly meteorological factors) have been applied to PM2.5 concentration prediction, revealing a clear discrepancy in timescale. To capture the complicated multi-scale relationship with PM2.5-related factors, a novel multi-factor & multi-scale method is proposed for PM2.5 forecasting. Three major steps are taken: (1) multi-factor analysis, to select predictive factors via statistical tests; (2) multi-scale analysis, to extract scale-aligned components via multivariate empirical mode decomposition; and (3) PM2.5 prediction, including individual prediction at each timescale and ensemble prediction across different timescales. The empirical study focuses on the PM2.5 of Cangzhou, which is one of the most air-polluted cities in China, and indicates that the proposed multi-factor & multi-scale learning paradigms statistically outperform their corresponding original techniques (without multi-factor and multi-scale analysis), semi-improved variants (with either multi-factor or multi-scale analysis), and similar counterparts (with other multi-scale analyses) in terms of prediction accuracy.

In recent years, the Chinese government has paid more and more attention to the construction of ecological civilization, and the governance of air pollution is one of the necessary links. Aiming at the problem of the time series prediction of air pollutants, this paper develops a time series prediction model based on deep learning method. In this paper, Beijing's hourly PM2.5 concentration information and weather information are used as input. Through GRU model, four models are trained according to the four seasons of spring, summer, autumn, and winter, and the effects of the four models on predicting the corresponding seasonal PM2.5 are evaluated by using corresponding test sets. After repeated experiments and constant adjustment of model parameters, the prediction error and prediction accuracy of the model are analyzed and compared, then the feasibility and advantages of this method are verified. The results delineate that the prediction accuracy of the model based on the GRU model is high, and the method is valid for the time series prediction of air pollutants.

Air samples of polycyclic aromatic hydrocarbons (PAHs) were collected from May 2008 to April 2009 in the Kaohsiung coastal area of Taiwan to establish a dataset. Sample analysis suggested that PAH composition (dominated by 3-ring PAHs) and seasonal patterns (high concentration in winter and low concentration in summer) are nearly identical to those previously reported (Lai et al., 2011), except for a case of road construction during the sample period. The predominant sources of PAHs were identified by isomer ratio, HCA and PCA analyses as vehicle emissions and coal/wood combustion. According to the results of parameters selection via the Group Method of Data Handling (GMDH), the variations in gaseous PAHs and particulate PAHs were related mainly to meteorological conditions and to routine monitoring of air pollutant concentrations, respectively. A comparison of model prediction accuracy between particulate and gaseous PAHs suggests better prediction for particulate PAHs than gaseous PAHs, due to dramatic variation in meteorological conditions which results in higher prediction uncertainty for gaseous PAHs. Model predictions of PAH concentrations showed satisfactory agreement with measured data, except for specific incidents, such as joss paper burning events. The results of this study suggest that the GMDH model may be applied to PAH concentration prediction for the study area and may be applicable for other countries based on routine monitoring of air pollutant concentrations and meteorological conditions.

The measurement for riser’s solids concentration is important but difficult. So, a high-efficiency and low-cost new predicted method based on the least square SVM method is proposed. The superficial operated gas velocity Ug and solids circulation rate εs are chosen as the effect factors of the solids concentration. Furthermore, the multiple linear regression model (Linear) and artificial neural net model (ANN) are also used to compare with the SVM model. By analyzing the training time and error, it is proved the SVM model has higher prediction accuracy and forecasting efficiency, which is a better choice for the riser’s solids concentration prediction.

In this paper, a novel framework and methodology based on hidden semi-Markov models (HSMMs) for high PM"2"."5 concentration value prediction is presented. Due to lack of explicit time structure and its short-term memory of past history, a standard hidden Markov model (HMM) has limited power in modeling the temporal structures of the prediction problems. To overcome the limitations of HMMs in prediction, we develop the HSMMs by adding the temporal structures into the HMMs and use them to predict the concentration levels of PM"2"."5. As a model-driven statistical learning method, HSMM assumes that both data and a mathematical model are available. In contrast to other data-driven statistical prediction models such as neural networks, a mathematical functional mapping between the parameters and the selected input variables can be established in HSMMs. In the proposed framework, states of HSMMs are used to represent the PM"2"."5 concentration levels. The model parameters are estimated through modified forward-backward training algorithm. The re-estimation formulae for model parameters are derived. The trained HSMMs can be used to predict high PM"2"."5 concentration levels. The validation of the proposed framework and methodology is carried out in real world applications: prediction of high PM"2"."5 concentrations at O'Hare airport in Chicago. The results show that the HSMMs provide accurate predictions of high PM"2"."5 concentration levels for the next 24h.

Plume concentration prediction is one of the main contents of radioactive consequence assessment for early emergency response to nuclear accidents. Random characteristics of atmospheric diffusion itself was described, a random walk model of atmospheric diffusion (Random Walk) was introduced and compared with the Lagrangian puff model (RIMPUFF) in the nuclear emergency decision support system (RODOS) developed by the European Community for verification. The results show the concentrations calculated by the two models are quite close except that the plume area calculated by Random Walk is a little smaller than that by RIMPUFF. The random walk model for atmospheric diffusion can simulate the atmospheric diffusion in case of nuclear accidents, and provide more actual information for early emergency and consequence assessment as one of the atmospheric diffusion module of the nuclear emergency decision support system.

Background: The osteocyte-derived hormone, fibroblast growth factor 23 (FGF23), regulates the phosphorus metabolism and suppresses 1,25-dihydroxyvitamin D production, thereby mitigating hyperphosphatemia in patients with renal disorders. An elevated FGF23 level is suggested to be an early biomarker of altered phosphorus metabolism in the initial stages of chronic kidney disease (CKD) and acts as a strong predictor of mortality in dialysis patients. In the Saudi population, there is no report on the FGF23 level in CKD patients to date. This study aims to estimate the plasma FGF23 levels in the Saudi population and to correlate it with its clinical manifestations in order to ascertain its role in the pathogenesis of CKD patients. Methods: The FGF23 level in the plasma samples was determined using ELISA in a diverse cohort of 89 cases with stage 3-5 CKD and 100 healthy subjects. The plasma FGF23 level was correlated with other biochemical parameters. Results: The results revealed that the FGF23 level was markedly elevated among CKD patients compared to the control group, and a significant inverse correlation was observed between the FGF23 level and glomerular filtration rate. FGF23 elevation was approximately 40-fold among stage 5 patients compared to the control, while the elevation of phosphate, parathyroid hormone (PTH) and alkaline phosphatase was 2-, 3- and 8-fold in this stage, respectively. Conclusion: Elevated FGF23 levels may have a strong correlation with the disease pathogenesis. In addition, FGF23 might be a future therapeutic target to intervene against the progression of CKD as well as to increase patient survivability.

One of the major concerns in Wastewater Treatment Plant (WWTP) operation is that of satisfying the legal requirements that impose maximum allowable concentration levels for effluent pollutants. Not meeting these requirements may generate economic punishment in terms of fines in addition, of course, to the environmental consequences. The effluent limit violations is usually measured as a side performance measure to existing WWTP control and operation approaches. However no explicit way of tackling this issue is found. In this paper a first step towards this direction is proposed in terms of a prognostication of the situations of risk. This is to say when the effluent is close to generate a limit violation for some of the limiting components. This is accomplished by means of effluent pollutants concentration prediction by using Artificial Neural Networks (ANN). The prediction is applied to a controlled plant and it is shown how a logical signal (therefore amenable for monitoring and decision) can be generated at the instants where such a risk is detected.

With decreasing water availability as a result of climate change and human activities, analysis of the influential factors and variation trends of chlorophyll a has become important to prevent reservoir eutrophication and ensure water supply safety. In this paper, a structurally simplified hybrid model of the genetic algorithm (GA) and the support vector machine (SVM) was developed for the prediction of monthly concentration of chlorophyll a in the Miyun Reservoir of northern China over the period from 2000 to 2010. Based on the influence factor analysis, the four most relevant influence factors of chlorophyll a (i.e., total phosphorus, total nitrogen, permanganate index, and reservoir storage) were extracted using the method of feature selection with the GA, which simplified the model structure, making it more practical and efficient for environmental management. The results showed that the developed simplified GA-SVM model could solve nonlinear problems of complex system, and was suitable for the simulation and prediction of chlorophyll a with better performance in accuracy and efficiency in the Miyun Reservoir.

Goal: For conventional modeling methods, the work of model identification has to be repeated with sufficient data for each subject because different subjects may have different response to exogenous inputs. This may cause repetitive cost and burden for patients and clinicians and require a lot of modeling efforts. Here, to overcome the aforementioned problems, a rapid model development strategy for new subjects is proposed using the idea of model migration for online glucose prediction. Methods: First, a base model is obtained that can be empirically identified from any subject or constructed by a priori knowledge. Then, parameters of inputs in the base model are properly revised based on a small amount of new data from new subjects so that the updated models can reflect the specific glucose dynamics excited by inputs for new subjects. These problems are investigated by developing autoregressive models with exogenous inputs (ARX) based on 30 in silico subjects using UVA/Padova metabolic simulator. Results: The prediction accuracy of the rapid modeling method is comparable to that for subject-dependent modeling method for some cases. Also, it can present better generalization ability. Conclusion: The proposed method can be regarded as an effective and economic modeling method instead of repetitive subject-dependent modeling method especially for lack of modeling data.

The rule-based approach to methane concentration prediction is presented in this paper. The applied solution is based on the modification called fixed of the separate-and-conquer rule induction approach. We also proposed the modification of a rule quality evaluation based on confidence intervals calculated for positive and negative examples covered by the rule. The characteristic feature of the considered methane forecasting model is that it omits the readings of the sensor being the subject of forecasting. The approach is evaluated on a real life data set acquired during a week in a coal mine. The results show the advantages of the introduced method (in terms of both the prediction accuracy and knowledge extraction) in comparison to the standard approaches typically implemented in the analytical tools.

Many spectroscopic monitoring techniques make use of chemometric algorithms such as principal component regression (PCR) for calibration and evaluation of optical spectra. However, standard techniques of chemometric calibration and evaluation suffer from uncalibrated spectral features which are contained in measured spectra but were not considered during the initial calibration. For two main reasons such uncalibrated spectral features have to be detected and corrected: (a) if unknown spectral features appear after calibration major errors of concentration prediction are encountered and (b) the detection of uncalibrated spectral features enables recognition and handling of failures occurring in chemical processes. Hence detection, correction and characterization of uncalibrated absorption features are of great importance in on-line spectroscopy. A novel approach is proposed here for this purpose. This algorithm performs a qualitative data analysis for detecting and correcting such uncalibrated spectral features. Following this qualitative data analysis and if necessary correction, a conventional PCR is used for calculating the concentration of the calibrated analytes. Both steps together set up a so-called secured PCR (sPCR). In the undisturbed case this sPCR algorithm is demonstrated to provide equivalent results as conventional PCR and to be superior in evaluating disturbed spectra. For testing purposes simulated spectra sets are used since the disturbance is known and the determined results can be validated with respect to the magnitude of the true disturbance. The errors averaged over 1000 disturbed test samples could be decreased up to 71% by sPCR compared with conventional PCR. In up to 94% of the test samples more accurate concentration values were obtained. The novel sPCR algorithm was also applied to experimental spectra measured with mid-infrared attenuated total reflection spectroscopy considering chlorinated hydrocarbons dissolved in water. sPCR is capable to extract uncalibrated features from these experimental data and to improve the concentration results. Copyright © 2003 John Wiley & Sons, Ltd.

In this work, vacuum deposited thin films of PbPc, NiPc, VOPc, TiOPc and CoPc were employed as gas sensor to detect NO2 and NO. Data collected from sensor responses were used to train a back-propagation network (BPN) for identifying the gas species and quantifying its concentration. The results show that among the metallophthalocyanines tested, PbPc and NiPc have better sensing characteristics towards NO2 and NO. In BPN training, maximum error occurs for data collected by the TiOPc sensor, and minimum error occurs for array of PbPc and NiPc sensors. In the concentration prediction of NO or NO2, the maximum predicted error is 6.94%. When Two-Stage BPN or Single-Stage BPN was use to identify and quantify a single gas (NO2 or NO), the accuracy of recognition approaches 100% and the maximum error for concentration prediction is 7.45%.

Hypoglycemia is a major challenge of artificial pancreas systems and a source of concern for potential users and parents of young children with Type 1 diabetes (T1D). Early alarms to warn the potential of hypoglycemia are essential and should provide enough time to take action to avoid hypoglycemia. Many alarm systems proposed in the literature are based on interpretation of recent trends in glucose values. In the present study, subject-specific recursive linear time series models are introduced as a better alternative to capture glucose variations and predict future blood glucose concentrations. These models are then used in hypoglycemia early alarm systems that notify patients to take action to prevent hypoglycemia before it happens. The models developed and the hypoglycemia alarm system are tested retrospectively using T1D subject data. A Savitzky-Golay filter and a Kalman filter are used to reduce noise in patient data. The hypoglycemia alarm algorithm is developed by using predictions of future glucose concentrations from recursive models. The modeling algorithm enables the dynamic adaptation of models to inter-/intra-subject variation and glycemic disturbances and provides satisfactory glucose concentration prediction with relatively small error. The alarm systems demonstrate good performance in prediction of hypoglycemia and ultimately in prevention of its occurrence.

We retrospectively evaluated the ability of equations with in vivo population binding parameters of our previous study (Method 1) or an average unbound fraction of 0.25 of Pynnönen (Method 2) to predict the unbound serum carbamazepine (CBZ) concentration in 35 serum samples from 18 adult patients with epilepsy receiving polytherapy. In 9 serum samples from 6 patients, the ability of equations for unbound serum carbamazepine-10,11-epoxide (CBZ-E) concentration prediction was also determined in predictive performance with in vivo population binding parameters of our previous study (Method A) or an average unbound fraction of 0.5 of Pynnönen (Method B). Mean prediction error, mean absolute prediction error (MAE), and root mean squared error (RMSE) were calculated for each method, and these values served as a measure of prediction bias and precision. Method 1 shows a bias to underpredict unbound serum CBZ. The MAE and RMSE were smaller in Method 2 (MAE = 0.454mgr;m/L, RMSE = 0.597mgr;m/L) than Method 1 (MAE = 0.597mgr;m/L, RMSE = 0.721mgr;m/L). Method 2 is superior to Method 1 in accuracy and precision. The effects of antiepileptic comedications on predictive performance of Methods 1 and 2 were determined in each group of serum samples with (n = 18, Group 1) or without (n = 17, Group 2) valproic acid (VPA) comedication. The results obtained by Method 1 show a bias to underprediction in Group 1 and no bias to over- or underprediction in Group 2. Results obtained by Method 2 show no bias to over- or underprediction in Groups 1 and 2. The effects of VPA comedication on predictive performance of unbound serum CBZ are relatively larger in Method 1 than Method 2. There was a weak but significant positive relationship between age and unbound serum CBZ fraction by simple regression analysis (n = 35, r = 0.368, p = 0.0297). The determined coefficient indicated that only about 14% of variations in unbound serum CBZ fraction can be explained by age, however. In each of Groups 1 and 2, no significant relationship was observed between age and unbound serum CBZ fraction. The effects of age on predictive performance of unbound serum CBZ are relatively small in patients receiving polytherapy. For unbound CBZ-E prediction, each of Methods A and B has no bias to over- or underprediction. The MAE was larger in Method B (MAE = 0.311mgr;m/L) than Method A (MAE = 0.233mgr;m/L). The differences in RMSE were small between Methods A (RMSE = 0.349mgr;m/L) and B (RMSE = 0.333mgr;m/L), however. Each of Methods A and B may have similar accuracy and precision.

A simplified building airflow model is presented that can be used to predict the spread of a contaminant agent from a chemical or biological attack. If the dominant means of agent transport throughout the building is an air-handling system operating at steady-state, a linear time-invariant (LTI) model can be constructed to predict the concentration in any room of the building as a result of either an internal or external release. While the model does not capture weather-driven and other temperature-driven effects, it is suitable for concentration predictions under average daily conditions. The model is easily constructed using information that should be accessible to a building manager, supplemented with assumptions based on building codes and standard air-handling system design practices. The results of the model are compared with a popular multi-zone model for a simple building and are demonstrated for building examples containing one or more air-handling systems. The model can be used for rapid concentration prediction to support low-cost placement strategies for chemical and biological detection sensors.

Rapid freeze prototyping is a solid freeform fabrication method that uses water freezing into ice as the build material. Each layer of geometry is deposited and allowed to freeze before the next layer is added in order to additively create a three-dimensional ice part. A sacrificial support material is needed for the fabrication of complex ice parts. Identifying a suitable support material and understanding the interaction between the build and support materials is the motivation behind this study. A temperature prediction model and a concentration prediction model are presented. Experimental results have been obtained to validate these models.

There are many substances in the air which may impair the health of plants and animals, including humans, that arise both from natural processes and human activity. Nitrogen dioxide NO 2 and particulate matter (PM 10 , PM 2.5 ) emissions constitute a major concern in urban areas pollution. The state of the air is, in fact, an important factor in the quality of life in the cities, since it affects the health of the community and directly influences the sustainability of our lifestyles and production methods. In this paper we propose a cyclostationary neural network (CNN) model for the prediction of the NO 2 and PM 10 concentrations. The cyclostationary nature of the problem guides the construction of the CNN architecture, which is composed by a number of MLP blocks equal to the cyclostationary period in the analyzed phenomenon, and is independent from exogenous inputs. Some experiments are also reported in order to show how the CNN model significantly outperforms standard statistical tools and linear regressors usually employed in these tasks.

Transformation products of environmental contaminants are likely to contribute significantly to the overall chemical pressure on valuable environmental resources. However, the whole extent of this so far often over-looked additional contamination remains unclear because the number of monitoring studies addressing transformation products is currently small and they are largely focused on well-known pesticide transformation products. Environmental fate modelling of transformation products opens up the possibility to predict the likely presence of environmental transformation products in environmental compartments of interest and to point towards new, potentially relevant transformation products. Whereas, depending on their purpose, there are different types of fate models for transformation products, this chapter will focus on multispecies multimedia models because of their general applicability to various tasks in chemical risk assessment and quality assessment of environmental resources. The chapter will introduce the mathematical framework underlying most multispecies multimedia models and discuss its application to three examples. These examples include the extension of overall persistence to include transformation products, the assessment of the long-range transport potential of persistent transformation products of semivolatile organic compounds, and the prediction of relative concentrations of pesticide transformation products in surface water bodies. The second part of the chapter discusses data requirements and availability for multispecies multimedia models and sheds some light on the accuracy of frequently used chemical property estimation tools. Lastly, tools to predict transformation schemes in those cases where no information on possible transformation products is available are introduced and their current limitations are discussed.

The concentration prediction of gas-in-oil for power transformer is an effective way to evaluate the transformer insulation on-line and prevent the transformer from further deterioration. However most conventional methods, such as GM(1,1) model consider each index separately. To make up for the disadvantage, the multivariable grey model (MGM(l,n)) is introduced to allow for the change in several different indices described simultaneously. The MGM(1,7) model for the seven characteristic gases monitored on-line in the power transformer oil is then developed. In the built model the interaction among these gases is comprehensively considered. The effectiveness and advantages of the proposed approach is demonstrated by the actual prediction case. (4 pages)

The capability of optical coherence tomography (OCT) to predict blood glucose concentration changes within physiological range was studied in vivo. Accuracy of 2 mM in prediction of blood glucose concentration by OCT has been achieved.

The sections in this article are Introduction Data Modeling by Latent Variables Decomposition of Spectral Data Using Latent Variables Least Squares Regression by Matrix Algebra “Partial” CLS to Yield a Principal Component Principal Component Regressions Comparison of CLS and ILS Generate Factors by PCA for ILS Protocols of PCR Calculation Steps of PCR Calibration Calculation Steps for Concentration Prediction by PCR Selection of Significant Factors Analysis of Error Empirically Defined Method of Error Estimation Statistical Method for Error Estimation Selection by Validation Abstract Spectra and Residuals Optional Pretreatment Mean Centering Scaling Normalization Partial Least Squares Regression PLS2 (Global PLS) PLS2 Calibration Discussion of the PLS2 Algorithm Selection of Basis Factors for PLS Modeling PLS Prediction PLS Modeling by Eigenvector Calculation PLS1 (Standard PLS) Comparison of PLS2 with PLS1 Comparison of PLS with PCR

Introduction We hypothesized that fuzzy logic could be used for pharmacokinetic modeling. Our objectives were to develop and evaluate a model for predicting serum lithium concentrations with fuzzy logic. Methods Steady-state pharmacokinetic data had been previously collected in 10 elderly patients (age range, 67 to 80 years) with depression who were receiving lithium once daily. Each patient had serial serum lithium concentration determinations over one 24-hour period. The resulting 137 data sets initially consisted of five input variables (age, weight, serum creatinine, lithium dose, and time since last dose) and one output variable (serum lithium concentration; range, 0.2 to 1.24 mmol/L). Results A fuzzy rulebase was created with 87 randomly chosen data sets, and predictions of serum lithium concentration were made on the basis of the remaining 50 data sets. All of the input variables except age and weight were identified as contributing to the fuzzy logic model. The average magnitude of the error in the predictions was 0.13 mmol/L (root mean squared error) with a bias (mean of the prediction errors) of 0.03 mmol/L. Conclusions This study indicates that the use of fuzzy logic for pharmacokinetic modeling of lithium for serum concentration predictions is feasible. Clinical Pharmacology & Therapeutics (1997) 62, 29–40; doi

A first step towards the multidetection, identification and quantification of anabolic androgenic steroids by enzyme-linked immunosorbent assay (ELISA) has been performed in this study. This proposal combines multiple competitive ELISA assays with different cross-reactivity profiles and multivariate data analysis techniques. Data have been analyzed by principal component analysis in conjunction with a novel K-nearest line classifier. This proposal allows simultaneous detection of up to four different steroids in the range of concentration from 0.1 to 316.2 nM with a total rate of 90.6% of correct detection, even in the presence of cross-reactivities. A methodology for concentration prediction is also presented with satisfactory results.

A statistical approach is about working through the historical data in a general way and finding guides to future behaviour. In the present study, forecasting of the criteria pollutants has been done using simple statistical techniques and attempt has been made for an inter-comparison of these techniques with various advanced statistical and deterministic techniques. The inter-comparison analysis leads us to the conclusion that there is no single modelling approach which generates optimum results. Considering the uncertainty and unavailability of most of the inputs of deterministic and advance statistical techniques, the methods adopted here are proposed to have great potential for air pollution forecasting.

A combined strategy of clustering and support vector regression (SVR) methods is proposed to predict Cyclosporine A (CyA) concentration in renal transplant recipients. Clustering combats the high variability and non-stationarity of the time series and reports knowledge gain in the problem. The SVR outperforms other classical neural networks.

In this work, a facile method is proposed to monitor the freshness of meat and fruits by combining a gas sensor array with machine learning algorithms, where the sensor array consists of four commercial metal oxide-based gas sensors. Back-propagation neural network (BPNN) is used for gas classification, and the average accuracy can reach 98.8%. To obtain more effective prediction of VOC gas concentration (ethanol, trimethylamine, and ammonia), four algorithms including BPNN, radial basis function neural network (RBFNN), support vector machine (SVM), and hybrid LDA-SVM, which is a combination of SVM and linear discriminant analysis (LDA) are implemented, which are trained with the same training set. By analyzing and comparing the prediction results of these four algorithm models, the RBFNN achieves the peak performance for the concentration predictions of ethanol and ammonia, and the average relative errors are less than 5% and 6.5%, respectively. For trimethylamine (TMA) concentration prediction, the average relative error of RBFNN is equal to 4.41%, which is better than 5.11% of SVM, while the mean absolute error of RBFNN is slightly inferior to SVM. Therefore, the classification accuracy of the gas type by BPNN and the prediction accuracy of gas concentration by RBFNN can meet the requirement of distinguishing the freshness of food.