Support vector regression with reduced training sets for air temperature prediction: a comparison with artificial neural networks

Sudden changes in weather, in particular extreme temperatures, can result in increased energy expenditures, depleted agricultural resources, and even loss of life. However, these ill effects can be reduced with accurate air temperature predictions that provide adequate advance warning. Support vector regression (SVR) was applied to meteorological data collected across the state of Georgia in order to produce short-term air temperature predictions. A method was proposed for reducing the number of training patterns of massively large data sets that does not require lengthy pre-processing of the data. This method was demonstrated on two large data sets: one containing 300,000 cold-weather training patterns collected during the winter months and one containing 1.25 million training patterns collected throughout the year. These patterns were used to produce predictions from 1 to 12 h ahead. The mean absolute error (MAE) for the evaluation set of winter-only patterns ranged from 0.514°C for the 1-h prediction horizon to 2.303°C for the 12-h prediction horizon. For the evaluation set of year-round patterns, the MAE ranged from 0.513°C for the 1-h prediction horizon to 1.922°C for the 12-h prediction horizon. These results were competitive with previously developed artificial neural network (ANN) models that were trained on the full data sets. For the winter-only evaluation data, the SVR models were slightly more accurate than the ANN models for all twelve of the prediction horizons. For the year-round evaluation data, the SVR models were slightly more accurate than the ANN models for three of the twelve prediction horizons.     

A genetic algorithm to refine input data selection for air temperature prediction using artificial neural networks

The accurate prediction of air temperature is important in many areas of decision-making including agricultural management, transportation and energy management. Previous research has focused on the development of artificial neural network (ANN) models to predict air temperature from one to twelve hours in advance. The inputs to these models included a constant duration of prior data with a fixed resolution for all environmental variables for all prediction horizons. The overall goal of this research was to develop more accurate ANN models that could predict air temperature for each prediction horizon. The specific objective was to determine if the ANN model accuracy could be improved by applying a genetic algorithm (GA) for each prediction horizon to determine the preferred duration and resolution of input prior data for each environmental variable. The ANN models created based on this GA based approach provided smaller errors than the models created based on the existing constant duration and fixed data resolution approach for all twelve prediction horizons. Except for a few cases, the GA generally included a longer duration for prior air temperature data and shorter durations for other environmental variables. The mean absolute errors (MAEs) for the evaluation input patterns of the one-, four-, eight-, and twelve-hour prediction models that were based on this GA approach were 0.564 °C, 1.264 °C, 1.766 °C and 2.018 °C, respectively. These MAEs were improvements of 3.98%, 4.59%, 2.55% and 1.70% compared to the models that were created based on the existing approach for the same corresponding prediction horizons. Thus, the GA based approach to determine the duration and resolution of prior input data resulted in more accurate ANN models than the existing ones for air temperature prediction. Future work could examine the effects of various GA and fitness evaluation parameters that were part of the approach used in this study.     

ENSEMBLE ARTIFICIAL NEURAL NETWORKS FOR PREDICTION OF DEW POINT TEMPERATURE

Dew point temperature is needed as an input to calculate various meteorological variables. In general, it contributes to human and animal comfort levels. The goal of this study was to develop artificial neural network (ANN) models for dew point temperature prediction to improve upon previous research. These improvements included optimizing the stopping criteria, comparing seasonal models to year-round models, and developing ensemble ANNs to blend the output of seasonal models. For an ANN trained with 100,000 patterns per epoch, the error was reduced using a 2000-pattern stopping dataset at an interval of 20 learning events to decide when to stop training. Seasonal ANN models were blended in an ensemble ANN with the weight of the member networks determined using a fuzzy membership-type function based on the day of year. These ensemble models were shown to produce lower errors than year-round, nonensemble models. The mean absolute errors (MAEs) of the final models evaluated with an independent evaluation dataset included 0.795°C for a 2-hour prediction, 1.485°C for a 6-hour prediction, and 2.146°C for a 12-hour prediction. The final model MAEs, when compared to the previous research, were reduced by 0.008°C, 0.081°C, and 0.135°C, respectively. It can be concluded that the methods used in this research were effective in more accurately predicting year-round dew point temperature. The ANN models for different prediction periods were sequenced to provide a 12-hour dew point temperature prediction system for implementation on the Georgia Automated Environmental Monitoring Network website (www.georgiaweather.net).     

Dew point and relative humidity measurement using a quartz resonant sensor

A new dew point measurement device for humidity measurement in high temperature environment using a quartz crystal sensor was proposed. Combined with Peltier module and quartz crystal, active condensation occurs in the quartz surface to change the mass on the surface of quartz crystal, and use the shift of its resonant frequency identify the time of condensation. This quartz sensor does not require any absorbent material, and it is directly stuck on the Peltier element. The sensor system can also be achieved relative humidity measurement based on dew point and ambient-temperature measuring. It can operate in the range of dew point temperature from 50 to ?30 °C, and in the range of relative humidity from 1 to 90 % RH. The measured dew points values and relative humidity values showed very good agreement with reference values and were within ±0.3 °C, 1 % RH, respectively over the whole temperature range.     

Prediction of suspended sediment concentration from water quality variables

This study investigates use of water quality (WQ) variables, namely total chromium concentration, total iron concentration, and turbidity for predicting suspended sediment concentration (SSC). For this purpose, the artificial neural networks (ANNs) and regression analysis (RA) models are employed. Seven different RA models are constructed, considering the functional relation between measured WQ variables and SSC. The WQ and SSC data are fortnightly obtained from six monitoring stations, located on the stream Harsit, Eastern Black Sea Basin, Turkey. A total of 132 water samples are collected from April 2009 to February 2010. Model prediction results reveal that ANN is able to predict SSC from WQ data, with mean absolute error (MAE) of 10.30 mg/L and root mean square error (RMSE) of 13.06 mg/L. Among seven RA models, the best one, which has the form including all independent parameters, produces results comparable to those of ANN, with MAE = 14.28 mg/L and RMSE = 15.35 mg/L. The sensitivity analysis results reveal that the most effective parameter on the SSC is total chromium concentration. These results have time- and cost-saving implications.     

Study on the estimation of near-surface air temperature from MODIS data by statistical methods

Spatially distributed air temperature is desired for various scientific studies, including climatalogical, hydrological, agricultural, environmental and ecological studies. In this study, empirical models with regard to land cover and spatial scale were introduced and compared to estimate air temperature from satellite-derived land surface temperature and other environmental parameters. Aqua MODIS (Moderate Resolution Imaging Spectroradiometer) data and meteorological data obtained throughout 2005 in the Yangtze River Delta were adopted to develop statistical algorithms of air temperature. Four empirical regression models with different forms and different independent variables resulted in errors ranging from 2.20°C to 2.34°C. Considering the different relationships between air temperature and land surface temperature for different land types, these four models were evaluated and the most proper equation for each land-cover type was determined. The model containing these selected equations gave a slightly improved mean absolute error (MAE) of 2.18°C. Then the spatial scale effect of this empirical model was analysed with observed air temperature and spatially averaged land surface characteristics. The result shows that the estimation error of air temperature tends to be lower as spatial window size increases, suggesting that the model performances are improved by spatially averaging land surface characteristics. Comprehensively considering the accuracy and computational demand, 5 × 5 pixel size is the most favourable window size for estimating air temperature. The validation of the empirical model at 5 × 5 pixel size shows that it achieves an MAE of 1.98°C and an R ² of 0.9215. This satisfactory result indicates that this approach is proper for estimating air temperature, and spatial window size is an important factor that should be considered when calculating air temperature. It is expected that better accuracy will be achieved if the different weights of pixels at different distances can be set according to high-density micro-meteorological data.     

Multiple Kernel Learning with Fisher Kernels for High Frequency Currency Prediction

Financially motivated kernels based on EURUSD currency data are constructed from limit order book volumes, commonly used technical analysis methods and canonical market microstructure models—the latter in the form of Fisher kernels. These kernels are used through their incorporation into support vector machines (SVM) to predict the direction of price movement for the currency over multiple time horizons. Multiple kernel learning is used to replicate the signal combination process that trading rules embody when they aggregate multiple sources of financial information. Significant outperformance relative to both the individual SVM and benchmarks is found, along with an indication of which features are the most informative for financial prediction tasks. An average accuracy of 55% is achieved when classifying the direction of price movement into one of three categories for a 200 s predictive time horizon.     

Numerical analysis on corrosion resistance of mild steel structures

Corrosion resistances of mild steel specimens according to artificial neural network (ANN) analysis were investigated in the scope of this study. Corrosion rate values were taken into numerical analysis as a result of experimental studies under corrosive aggressive media. Mild steel specimens were selected according to the section type varieties such as box, tube and cornier. All steel specimens were subjected to the aggressive media formed using sodium chloride (NaCl with 99.8 % purity) solutions with 3.5, 5.0 and 7.0 % ratios per one liter distilled water and only distilled water. The reduction in corrosion rate has been observed and considered according to some corrosion loss respects. Corrosion rate prediction models were established between corrosion rate and parameters such as mass loss obtained by experimental studies using ANN. ANNs are computing systems that simulate the biological neural systems of the human brain. In this study, ANN analysis was generated to predict the corrosion rate values after experimental studies. Experimental and predicted values were compared by each other and it is seen that a strong relationship was established between them.     

Development of local ambient gas control technologies for atmospheric MEMS process

This paper reports a local ambient gas control technology for atmospheric MEMS processes, especially plasma processes, using a new local ambient gas control head. First, the local ambient gas control with this head was investigated by a computational fluid dynamics code. After confirmation of the safe evacuation and the feasible cleanness level, which is comparable to the impurity level in semiconductor grade gas (below 10 ppm), a prototype apparatus was fabricated based on the simulation results. Measuring gas distribution by a gas analyzer, a O₂ meter and a dew point meter, the local ambient gas control was confirmed experimentally. Next, H₂ plasma generation was achieved in open air with H₂ concentrations of 0–100 % even above the explosive limit in air (4.1 %) safely. In addition, Cu reduction and SiO₂ etching by H₂ plasma were demonstrated in open air. These results show high potential of our local ambient gas control technology for atmospheric MEMS processes.     

Flood flow forecasting using ANN,ANFIS and regression models

Flood prediction is an important for the design, planning and management of water resources systems. This study presents the use of artificial neural networks (ANN), adaptive neuro-fuzzy inference systems (ANFIS), multiple linear regression (MLR) and multiple nonlinear regression (MNLR) for forecasting maximum daily flow at the outlet of the Khosrow Shirin watershed, located in the Fars Province of Iran. Precipitation data from four meteorological stations were used to develop a multilayer perceptron topology model. Input vectors for simulations included the original precipitation data, an area-weighted average precipitation and antecedent flows with one- and two-day time lags. Performances of the models were evaluated with the RMSE and the R ². The results showed that the area-weighted precipitation as an input to ANNs and MNLR and the spatially distributed precipitation input to ANFIS and MLR lead to more accurate predictions (e.g., in ANNs up to 2.0 m³ s^?1 reduction in RMSE). Overall, the MNLR was shown to be superior (R ² = 0.81 and RMSE = 0.145 m³ s^?1) to ANNs, ANFIS and MLR for prediction of maximum daily flow. Furthermore, models including antecedent flow with one- and two-day time lags significantly improve flow prediction. We conclude that nonlinear regression can be applied as a simple method for predicting the maximum daily flow.     

Predicting river daily flow using wavelet-artificial neural networks based on regression analyses in comparison with artificial neural networks and support vector machine models

This study investigates the ability of wavelet-artificial neural networks (WANN) for the prediction of short-term daily river flow. The WANN model is improved by conjunction of two methods, discrete wavelet transform and artificial neural networks (ANN) based on regression analyses, respectively. The proposed WANN models are applied to the daily flow data of Vanyar station, on the Ajichai River in the northwest region of Iran, and compared with the ANN and support vector machine (SVM) techniques. Mean square error (MSE), mean absolute error (MAE) and correlation coefficient (R) statistics are used for evaluating precision of the WANN, ANN and SVM models. Comparison results demonstrate that the WANN model performs better than the ANN and SVM models in short-term (1-, 2- and 3-day ahead) daily river flow prediction.

     

High granular and short term time series forecasting of $$\hbox {PM}_{2.5}$$ PM 2.5 air pollutant - a comparative review

Forecasting time series has acquired immense research importance and has vast applications in the area of air pollution monitoring. This work attempts to investigate the abilities of various existing techniques when applied for short term, high granular time series forecasting of PM_2.5. More specifically, a comparative study has been provided, taking into account both popularly used models and lesser-used models in this area. The study has been carried out considering ten well defined models that are ARIMA (auto-regressive integrated moving average), SARIMA (seasonal ARIMA), SES (single exponential smoothing), DES (double exponential smoothing), TES (triple exponential smoothing), ANN (artificial neural network), DT (decision tree), kNN (k-nearest neighbor), LSTM (long short-term memory) and MCFO (markov chain first order). A framework has been built that categories the models, implements them under identical execution environment and forecasts succeeding values. Implementation has been carried out over five data sets of real-world air pollution time series, that are collected from five differently located government setup monitoring stations over a period of 1 year (July 2018-June 2019). Rigorous statistical analysis has been performed that yields an insight to the nature and variability of these time series data. Forecasting has been carried out on short term basis, focusing on high granularity whereas, three different lengths of forecast horizon (1 day, 1 week, and 1 month) have been tested. Eventually, the models have been compared in terms of their associated performance measuring units namely, RMSE (root mean of squared error), MAE (mean absolute error) and MAPE (mean absolute percentage error). The comparative results verified with multiple datasets show that all the models posses less error for a shorter forecast horizon, where LSTM providing the best performance. Superiority of machine learning and deep learning models are found in case of longer length of forecast horizon with kNN achieving best accuracy whereas, significant performance degradation of ARIMA is found for longer forecast horizon. Moreover, TES, DT, kNN, LSTM, MCFO are found to be well adopted in relation with shape and variability of the data. Note that the performance on various length of high granular forecast horizon have been studied over multiple datasets that give an added value to this work.

     

Neural network nonlinear modeling for hydrogen production using anaerobic fermentation

The potential of utilizing artificial neural network (ANN) model approach for simulate and predict the hydrogen yield in batch model using Clostridium saccharoperbutylacetonicum N1-4 (ATCC 13564) was investigated. A unique architecture has been introduced in this research to mimic the inter-relationship between three input parameters initial substrate, initial medium pH and reaction temperature (37 °C, 6.0 ± 0.2, 10), respectively, to predict hydrogen yield. Sixty data records from the experiment have been utilized to develop the ANN model. The results showed that the proposed ANN model provided significant level of accuracy for prediction with maximum error (10 %). Furthermore, a comparative analysis with a traditional approach Box–Wilson design (BWD) has proved that the ANN model output significantly outperformed the BWD. ANN model overcomes the limitation of the BWD approach with respect to the number of records, which is merely considering limited length of stochastic pattern for hydrogen yield (15 records).     

Customer churn prediction by hybrid neural networks

As churn management is a major task for companies to retain valuable customers, the ability to predict customer churn is necessary. In literature, neural networks have shown their applicability to churn prediction. On the other hand, hybrid data mining techniques by combining two or more techniques have been proved to provide better performances than many single techniques over a number of different domain problems. This paper considers two hybrid models by combining two different neural network techniques for churn prediction, which are back-propagation artificial neural networks (ANN) and self-organizing maps (SOM). The hybrid models are ANN combined with ANN (ANN + ANN) and SOM combined with ANN (SOM + ANN). In particular, the first technique of the two hybrid models performs the data reduction task by filtering out unrepresentative training data. Then, the outputs as representative data are used to create the prediction model based on the second technique. To evaluate the performance of these models, three different kinds of testing sets are considered. They are the general testing set and two fuzzy testing sets based on the filtered out data by the first technique of the two hybrid models, i.e. ANN and SOM, respectively. The experimental results show that the two hybrid models outperform the single neural network baseline model in terms of prediction accuracy and Types I and II errors over the three kinds of testing sets. In addition, the ANN + ANN hybrid model significantly performs better than the SOM + ANN hybrid model and the ANN baseline model.     

Forecasting highway casualties under the effect of railway development policy in Turkey using artificial neural networks

This study presents forecast of highway casualties in Turkey using nonlinear multiple regression (NLMR) and artificial neural network (ANN) approaches. Also, the effect of railway development on highway safety using ANN models was evaluated. Two separate NLMR and ANN models for forecasting the number of accidents (A) and injuries (I) were developed using 27 years of historical data (1980–2006). The first 23 years data were used for training, while the remaining data were utilized for testing. The model parameters include gross national product per capita (GNP-C), numbers of vehicles per thousand people (V-TP), and percentage of highways, railways, and airways usages (TSUP-H, TSUP-R, and TSUP-A, respectively). In the ANN models development, the sigmoid and linear activation functions were employed with feed-forward back propagation algorithm. The performances of the developed NLMR and ANN models were evaluated by means of error measurements including mean absolute percentage error (MAPE), mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R ²). ANN models were used for future estimates because NLMR models produced unreasonably decreasing projections. The number of road accidents and as well as injuries was forecasted until 2020 via different possible scenarios based on (1) taking TSUPs at their current trends with no change in the national transport policy at present, and (2) shifting passenger traffic from highway to railway at given percentages but leaving airway traffic with its current trend. The model results indicate that shifting passenger traffic from the highway system to railway system resulted in a significant decrease on highway casualties in Turkey.     

Artificial neural network–genetic algorithm for estimation of crop evapotranspiration in a semi-arid region of Iran

This study compares the daily potato crop evapotranspiration (ET_C) estimated by artificial neural network (ANN), neural network–genetic algorithm (NNGA) and multivariate nonlinear regression (MNLR) methods. Using a 6-year (2000–2005) daily meteorological data recorded at Tabriz synoptic station and the Penman–Monteith FAO 56 standard approach (PMF-56), the daily ET_C was determined during the growing season (April–September). Air temperature, wind speed at 2 m height, net solar radiation, air pressure, relative humidity and crop coefficient for every day of the growing season were selected as the input of ANN models. In this study, the genetic algorithm was applied for optimization of the parameters used in ANN approach. It was found that the optimization of the ANN parameters did not improve the performance of ANN method. The results indicated that MNLR, ANN and NNGA methods were able to predict potato ET_C at desirable level of accuracy. However, the MNLR method with highest coefficient of determination (R ² > 0.96, P value < 0.05) and minimum errors provided superior performance among the other methods.     

Artificial neural network for prediction of air flow in a single rock joint

In this paper, an attempt has been made to evaluate and predict the air flow rate in triaxial conditions at various confining pressures incorporating cell pressure, air inlet pressure, and air outlet pressure using artificial neural network (ANN) technique. A three-layer feed forward back propagation neural network having 3-7-1 architecture network was trained using 37 data sets measured from laboratory investigation. Ten new data sets were used for the, validation and comparison of the air flow rate by ANN and multi-variate regression analysis (MVRA) to develop more confidence on the proposed method. Results were compared based on coefficient of determination (CoD) and mean absolute error (MAE) between measured and predicted values of air flow rate. It was found that CoD between measured and predicted air flow rate was 0.995 and 0.758 by ANN and MVRA, respectively, whereas MAE was 0.0413 and 0.1876.     

基于t分布的贝叶斯深度学习模型及其应用

贝叶斯深度学习(BDL)融合了贝叶斯方法与深度学习(DL)的互补优势,成为复杂问题中不确定性建模与推断的强大工具.本文构建了基于t分布和循环随机梯度汉密尔顿蒙特卡罗采样算法的BDL框架,并基于数据不确定性和模型定不确定性给出了不确定性的度量.为了验证模型框架的有效性和适用性,我们分别基于人工神经网络(ANN)、卷积神经网络(CNN)和循环神经网络(RNN)构建了相应的BDL模型,并将模型应用于全球15个股票指数预测,实证结果显示:1)该框架在ANN、CNN和RNN下均适用,对全部指数的预测效果均很出色; 2)在预测精度和通用性方面,基于t分布BDL的模型比基于正态分布的BDL模型具有显著优越性; 3)在给定不确定性阈值之下的预测MAE比初始MAE显著提升,表明文中定义的不确定性是有效的,对不确定性建模具有重要意义.鉴于该BDL框架在预测精度、易于拓展和具备提供预测不确定性度量的优势,其在金融和其他具有复杂数据特征的领域均有广阔的应用前景.     

基于KNN-LSTM的PM2.5浓度预测模型

目前多数PM_2.5浓度预测模型仅利用单个站点的时间序列数据进行浓度预测, 并没有考虑到空气质量监测站之间的区域关联性, 这会导致预测存在一定的片面性. 本文利用KNN算法选择目标站点所在区域中与其相关的空间因素, 并结合LSTM模型, 提出基于时空特征的KNN-LSTM的PM_2.5浓度预测模型. 以哈尔滨市10个空气质量监测站的污染物数据进行仿真实验, 并将KNN-LSTM模型与其他预测模型进行对比, 结果显示: 模型相较于BP神经网络模型平均绝对误差(MAE)、均方根误差(RMSE)分别降低了19.25%、13.23%; 相较于LSTM模型MAE、RMSE分别降低了4.29%、6.99%. 表明本文所提KNN-LSTM模型能有效提高LSTM模型的预测精度.