Predicting lattice constant of complex cubic perovskites using computational intelligence

The Seventh International Symposium on Neural Networks (ISNN 2010) was held on June 6–9, 2010 in Shanghai, China. The ISNN 2010 was a great success and provided a high-level international forum for scientists and engineers to present the latest research in neural networks, computational intelligence, cognitive systems, and related fields. To highlight the success of this conference, we edited this special issue for Cognitive Computation. We chose nine papers from over 591 papers submitted to ISNN 2010, which reflect the high quality of the presentations at the conference while capturing the spirit of our theme, ‘‘Advances in Computational Intelligence and Applications,’’ for this special issue. Our goal for this special issue is to present the latest research developments in computational intelligence with a focus on cognitive computation and their applications across a wide range of domains. Over the past decades, we have witnessed tremendous interest and developments in all aspects of computational intelligence research, ranging from both theoretical foundations, principles, to practical applications in different domains. To reflect a flavor of recent research activities in the community, we carefully selected these nine papers for this special issue. The selected papers can be organized into the following four coherent sections. The first section is directly related to neural network models and applications. The paper by Deng presents a new control strategy for the chaotic neural network, in which the refractoriness is tuned by using feedback control based on online averaging of network states. Simulation results demonstrate that the proposed model can achieve favorable performance in handling noisy, incomplete, and composite patterns, while also achieving either enhanced or comparable memory capacity compared with the classic Hopfield net and other chaotic neural network models. In the other paper, Gluge et al. discuss recurrent networks for implicit learning of temporal sequences, a very important area to understand high-level cognitive intelligence. The authors propose a recurrent network with backpropagation training algorithm that is adapted to the reinforcement learning scheme. The simulation results fit quantitatively as well as qualitatively to the behavioral results, suggesting the role of temporal context in associative learning scenarios. The second section is related to language processing and understanding, an important topic in cognitive science and computational intelligence. The paper by Murata et al. investigates the maximum entropy (ME) approach for natural language processing (NLP) problems including machine translation and information extraction. Comparative studies of the proposed approach with the existing techniques are illustrated in this paper. In related work to this field, the paper by Kaznatcheev investigates connectionist models on the interplay of nouns and pronouns in personal pronoun acquisition, a very interesting and important problem in psychology. For instance, the author discusses the learning in the shifting reference situation for children. It is reported that learning of two different noun-and-pronoun addressee patterns is consistent with Z. Zeng (&) Department of Control Science and Engineering, Huazhong University of Science and Technology, Wuhan 430074, P. R. China e-mail: zgzeng@gmail.com

Stream-flow forecasting using extreme learning machines: A case study in a semi-arid region in Iraq

The aim of this work was to develop an accurate model to swiftly forecast the ultimate conditions and the stress–strain curves of ultra-high-performance concrete (UHPC) confined by fibre-reinforced polymer (FRP). Four neural network predictive models based on machine learning (back-propagation neural network (BPNN), support vector regression (SVR), extreme learning machine (ELM) and generalised regression neural network (GRNN)) were constructed to forecast the ultimate conditions. A BPNN-based model to predict stress–strain curves was developed. A database of 193 compression test results of FRP-confined UHPC members was collected from the open literature and a thorough evaluation of data quality was conducted. The results showed that the SVR, BPNN, GRNN and ELM prediction models exhibited excellent accuracy when compared with a design-oriented model. The GRNN exhibited the highest predictive precision, followed by SVR. The developed prediction models establish a foundation for the design and quick prediction of FRP-confined UHPC.

Monthly revenue per unit (RPU), measured by the total monthly revenue divided by the total monthly number of cars in the fleet, is one of the most popular profitability assessments in the car-rental industry. Because pricing decision is influenced by profit a lot, accurate forecast of RPU is thus useful for pricing in car-rental industry. Support vector regression (SVR) is an emerging forecast technique and successfully applied in many domains. The aim of this study is to investigate the feasibility of SVR in RPU forecast. Because SVR model contains three parameters influencing the forecasting accuracy, this work treats the forecasting error as an objective and employs tabu search (TS) algorithms to minimize the forecasting error. Actual RPU data are used to depict the forecasting performance of the improved SVR model. Two forecasting model, namely the general regression neural network (GRNN) model and the autoregressive integrated moving average (ARIMA) model, are used to forecast the same RPU data. Simulation results indicate that the improved SVR model is superior to the other models in terms of forecasting accuracy.

Development and analyses of data-driven models for predicting the bed depth profile of solids flowing in a rotary kiln

A computational intelligence (CI)-based approach is presented for prognostics of machine conditions using morphological signal processing (MSP). The machine vibration signals are processed using MSP to extract a novel entropy-based health index (HI) characterizing the signal shape-size complexity for system prognostics. The progression of HI is predicted using CI techniques, namely, recursive neural network (RNN), adaptive neuro-fuzzy inference system (ANFIS), and support vector regression (SVR). Both single- and multi-step ahead predictions were evaluated through benchmark datasets of non-linear, non-stationary, and chaotic time series solutions of Mackey-Glass and Lorenz equations. The prognostic effectiveness of the CI techniques was illustrated using a vibration dataset of a helicopter drive-train system gearbox. For each CI predictor, both training datasets gave almost similar prediction performance. In training, the performance of ANFIS was the best, followed by SVR and RNN. In test, the best performance was obtained with SVR for both single- and multi-step ahead predictions. The results are helpful in understanding the relationship between the system conditions, the corresponding indicating feature, the level of degradation, and their progression.

Leveraging hybrid machine learning and data fusion for accurate mapping of malaria cases using meteorological variables in western India

Classification of cancer patients into treatment groups is essential for appropriate diagnosis to increase survival. Previously, a series of papers, largely published in the breast cancer domain have leveraged Computational Intelligence (CI) developments and tools, resulting in ground breaking advances such as the classification of cancer into newly identified classes — leading to improved treatment options. However, the current literature on the use of CI to achieve this is fragmented, making further advances challenging. This paper captures developments in this area so far, with the goal to establish a clear, step-by-step pipeline for cancer subtype identification. Based on establishing the pipeline, the paper identifies key potential advances in CI at the individual steps, thus establishing a roadmap for future research. As such, it is the aim of the paper to engage the CI community to address the research challenges and leverage the strong potential of CI in this important area. Finally, we present a small set of recent findings on the Nottingham Tenovus Primary Breast Carcinoma Series enabling the classification of a higher number of patients into one of the identified breast cancer groups, and introduce Classifusion: a combination of results of multiple classifiers.

To quickly and accurately assess tea plant growth, this study aims to find a new way to predict the chlorophyll content in tea plant canopies using machine learning. Using remotely piloted aircraft equipped with multispectral cameras, images of tea plantation areas are captured and reflectance from four spectral bands is extracted, leading to the calculation of vegetation indices. Simultaneously, chlorophyll relative content in the tea plant canopies was collected on the ground using a detector. Four models, namely Random Forest (RF), Backpropagation neural network (BPNN), Radial basis function network (RBFN), and General Regression Neural Network (GRNN), were constructed to predict the chlorophyll relative content in tea plant canopies. Subsequently, important remote sensing variables were identified through RF filtering, followed by a comparison of the predictive performance of machine learning models under different input conditions. Lastly, by integrating the Sparrow Search Algorithm (SSA) to optimize the smoothing factor in the GRNN, the study explores the impact of optimization algorithms on the predictive performance of the GRNN model. Experiments indicate that within the established machine learning models, the GRNN demonstrates the highest predictive accuracy. By ranking the importance of remote sensing variables through RF, 18 significant remote sensing variables were selected, which enhanced the predictive accuracy of the machine learning models. The optimization of the GRNN smoothing factor through the SSA algorithm can significantly enhance the predictive accuracy of the GRNN model. Based on a series of experiments, the established RFSSA-GRNN prediction model demonstrates good predictive performance, with an reaching 0.84.

Beam pumping system which is widely used in petroleum enterprises of China is one of the most energy-consuming equipment. It is difficult to be modeled and optimized due to its complication and nonlinearity. To address this issue, a novel intelligent computing based method is proposed in this paper. It firstly employs the general regression neural network (GRNN) algorithm to obtain the best model of the beam pumping system, and secondly searches the optimal operation parameters with improved strength Pareto evolutionary algorithm (SPEA2). The inputs of GRNN include the number of punching, the maximum load, the minimum load, the effective stroke, and the computational pump efficiency, while the outputs are the electric power consumption and the oil yield. Experimental results show that there is good overlap between model estimations and unseen data. Then sixty-one sets of optimum parameters are found based on the obtained model. Also, the results show that, under the optimum parameters, more than 5.34% oil yield is obtained and more than 3.75% of electric power consumption is saved.

Accurate forecast of carbon dioxide (CO2) emissions plays a significant role in China's carbon peaking and carbon neutrality policies. A novel two-stage forecast procedure based on support vector regression (SVR), random forest (RF), ridge regression (Ridge), and artificial neural network (ANN) is proposed and evaluated by comparing it with the single-stage forecast procedure. Nine independent variables’ data (study period: 1985–2020) are used to forecast the CO2 emissions in China. Our results reveal that, when the time gap, h increases from 1 to 8, the average root mean squared error (RMSE) and mean absolute error (MAE) of SVR–SVR, SVR–RF, SVR–Ridge, and SVR–ANN are almost uniformly lower than errors arising from their single-stage version, respectively. Among these two-stage models, SVR–ANN exhibits the lowest forecast errors, whereas SVR–RF admits the highest. The mean percentage decrease in forecast errors of SVR–SVR vs. SVR, SVR–RF vs. RF, SVR–Ridge vs. Ridge, and SVR–ANN vs. ANN are 36.06, 5.98, 43.05, and 14.81 for RMSE, and 36.06, 6.91, 43.27, and 15.35 for MAE. Our two-stage procedure is also suitable to forecast other variables, such as fossil fuel and renewable energy consumption.

Soil CO2 emission (FCO2) is a critical component of the global carbon cycle, but it is a source of great uncertainty due to the great spatial and temporal variability. Modeling of soil respiration can strongly contribute to reducing the uncertainties associated with the sources and sinks of carbon in the soil. In this study, we compared five machine learning (ML) models to predict the spatiotemporal variability of FCO2 in three reforested areas: eucalyptus (RE), pine (RP) and native species (RNS). The study also included a generalized scenario (GS) where all the data from RE, RP and RNS were included in one dataset. The ML models include generalized regression neural network (GRNN), radial basis function neural network (RBFNN), multilayer perceptron neural network (MLPNN), adaptive neuro-fuzzy inference system (ANFIS) and random forest (RF). Initially, we had 32 attributes and after pre-processing, including Pearson's correlation, canonical correlation analysis (CCA), and biophysical justification, only 21 variables remained. We used as input variables 19 soil properties and climate variables in reforested areas of eucalyptus, pine and native species. RF was the best model to predict soil respiration to RE [adjusted coefficient of determination (R2 adj): 0.70 and root mean square error (RMSE): 1.02µmolm-2s-1], RP (R2 adj: 0.48 and RMSE: 1.07µmolm-2s-1) and GS (R2 adj: 0.70 and RMSE: 1.05µmolm-2s-1). Our findings support that RF and GRNN are promising for predicting soil respiration of reforested areas which could help to identify and monitor potential sources and sinks of the main additional greenhouse gas over ecosystems.

Computational Intelligence has made a huge impact on solving many complicated problem particularly in the medical field. With the advancement of computational intelligence where the effectiveness of data analysis is at high stake, the process of classifying and interpreting data accurately based on logical reasoning in decision making is not a big issue. This study discusses the process of diagnosing cardiac disorder using computational intelligence with specific focus on the feature extraction where the attribute of identifying Normal Sinus and Atrial Fibrillation rhythms using Physionet.org database is examined. In this paper, an algorithm to diagnose the cardiac disorder based on DP-Matching will be proposed where the time and frequency domains of ECG signal segments are introduced. At the end of this paper, the performance evaluations of the proposed method will be shown with the analysis by ANN.

Long-term wind power forecasting is a challenging endeavor that requires predictions that span years into the future. Accurate forecasting is crucial for optimizing energy production, grid integration, maintenance scheduling, and financial planning. This study attempts to first develop the long short-term memory networks (LSTM) with a seasonal wavelet transform forecasting model for practical long-term wind power forecasting problems with seasonal and regional influences on wind power and the instability of data signals. This model encapsulates wavelet transformation and seasonal decomposition, and employs LSTM for forecasting. The new prediction model adopted seasonal decompositions and two LSTMs to approach low- and high-frequency series datasets, as well as the wavelet synthesis prediction values. Furthermore, the parameters of the LSTM models are selected using stochastic optimization. For a comprehensive evaluation, the proposed LSTM with seasonal wavelet transform is compared with seven methods, including seasonal LSTM (SLSTM), wavelet LSTM (WLSTM), and the seasonal auto-regressive integrated moving average (SARIMA), back propagation neural network (BPNN), generalized regression neural network (GRNN), least square support vector regression (LSSVR), and support vector regression (SVR) were employed for long-term wind power output forecasting of wind farms. The empirical results underscore that the performance of the proposed forecasting model is better than other methods in terms of forecasting accuracy, which could efficiently provide reliable long-term predictions for long-term wind power output forecasting.

Electricity load demand is the fundamental building block for all utilities planning. The load demand data has nonlinear and non-stationary characteristics, which make it difficult to be predicted accurately by just computational intelligence or ensemble methods. Ensemble methods like Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) is a powerful tool to forecast power load demand time series. Heterogeneous ensemble, a combination of two base models, will be distinct or more powerful in forecasting power load demand. In this paper, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) method is hybridized with three computational intelligence-based predictors: support vector regression (SVR), artificial neural network (ANN) and random forest (RF). The basis of this paper was to conduct a comparative study on the accuracy of the forecasting result from using heterogeneous ensemble method to individual computational intelligence or ensemble method for four different horizons. The performances of the heterogeneous method are compared and discussed. It shows that heterogeneous method has outperformed the individual computational intelligence and ensemble methods. Possible future works are also recommended for power load demand forecasting.