A spectral envelope approach towards effective SVM-RFE on infrared data

Intelligent multivariable air-quality forecasting system based on feature selection and modified evolving interval type-2 quantum fuzzy neural network

Owing to the high nonlinearity and noise in the air quality index (AQI), tackling the uncertainties and fuzziness in the forecasting process is still a prevalent problem. Therefore, this study developed an intelligent hybrid air-quality forecasting system based on feature selection and a modified evolving interval type-2 quantum fuzzy neural network (eIT2QFNN), which provides accurate air-quality forecasting information by considering climate influencing factors. The main contributions of this study are as follows. The optimal input structure of the model is determined by the proposed second-stage feature-selection model, which can better extract the influencing variables and remove redundant information. Moreover, a novel multi-objective chaotic Bonobo optimizer algorithm is proposed to improve the eIT2QFNN. The modified eIT2QFNN implements AQI prediction by considering the importance of influencing variables that can cope with the uncertainties and fuzziness in the forecasting process. Finally, the Diebold-Mariano and modified Diebold-Mariano tests are employed to evaluate the performance of the proposed system. The experimental results demonstrate that our proposed system significantly improves the modeling performance in terms of high accuracy and compact structure, and can thus serve as an effective tool for air-quality management.

Detection and Classification of Bearing Surface Defects Based on Machine Vision

Surface defects on bearings can directly affect the service life and reduce the performance of equipment. At present, the detection of bearing surface defects is mostly done manually, which is labor-intensive and results in poor stability. To improve the inspection speed and the defect recognition rate, we proposed a bearing surface defect detection and classification method using machine vision technology. The method makes two main contributions. It proposes a local multi-neural network (Lc-MNN) image segmentation algorithm with the wavelet transform as the classification feature. The precision segmentation of the defect image is accomplished in three steps: wavelet feature extraction, Lc-MNN region division, and Lc-MNN classification. It also proposes a feature selection algorithm (SCV) that makes comprehensive use of scalar feature selection, correlation analysis, and vector feature selection to first remove similar features through correlation analysis, further screen the results with a scalar feature selection algorithm, and finally select the classification features using a feature vector selection algorithm. Using 600 test samples with three types of defect in the experiment, an identification rate of 99.5% was achieved without the need for large-scale calculation. The comparison tests indicated that the proposed method can achieve efficient feature selection and defect classification.

Effects of caprolactam content on curdlan-based food packaging film and detection by infrared spectroscopy

In this study, we report a rapid statistical approach used in determining the caprolactam (CPL) content in curdlan packaging films, which is based on the spectral data observed in the near-infrared (NIR) and Mid-infrared (MIR) regions. At the first stage of the study, the CPL content was added into the curdlan films prepared by controlling the concentration, and then the effect of the CPL concentration on the measured mechanical properties of the produced films were evaluated. At the next stage, the NIR and MIR spectra of the curdlan films with different CPL concentrations were recorded by using the FT-NIR and FT-IR spectroscopy technique, and the spectral data to be used in the regression models in our quantitative analyses were carefully selected. It was observed that the curdlan film with 5% CPL exhibited the best mechanical properties. The obtained best correlation parameters which are used in evaluation of CPL content through the observed NIR and MIR spectral data are Rp = 0.9552, RMSEP = 1.2506 (NIR); Rp = 0.9092 and RMSEP = 1.9136 (MIR), respectively. These optimal values support the expectation that our statistical approach based on NIR and MIR data can provide a rapid, accurate and nondestructive way of determining CPL content in curdlan packaging films.

Kernel functions embedded in support vector machine learning models for rapid water pollution assessment via near-infrared spectroscopy

Water pollution is a challenging problem encountered in total environmental development. Near-infrared (NIR) spectroscopy is a well-refined technology for rapid water pollution detection. Calibration models are established and optimized to search for chemometric algorithms with considerably improved prediction effects. Machine learning improves the prediction capability of NIR spectroscopy for the accurate assessment of water pollution. Least squares support vector machine (LSSVM) algorithm fits parameters to target problems in a data-driven manner. The modeling capability of this algorithm mainly depends on its kernel functions. In this study, the LSSVM method was used to establish NIR calibration models for the quantitative determination of chemical oxygen demand, which is a critical indicator of water pollution level. The effects of different kernels embedded in LSSVM were investigated. A novel kernel was proposed by using a logistic-based neural network. In contrast to common kernels, this novel kernel can utilize a deep learning approach for parameter optimization. The proposed kernel also strengthens model resistance to over-fitting such that cross-validation can be reasonably utilized. The proposed novel kernel is applicable for the quantitative determination of water pollution and is a prospective solution to other problems in the field of water resource management.

Decision Variants for the Automatic Determination of Optimal Feature Subset in RF-RFE

Feature selection, which identifies a set of most informative features from the original feature space, has been widely used to simplify the predictor. Recursive feature elimination (RFE), as one of the most popular feature selection approaches, is effective in data dimension reduction and efficiency increase. A ranking of features, as well as candidate subsets with the corresponding accuracy, is produced through RFE. The subset with highest accuracy (HA) or a preset number of features (PreNum) are often used as the final subset. However, this may lead to a large number of features being selected, or if there is no prior knowledge about this preset number, it is often ambiguous and subjective regarding final subset selection. A proper decision variant is in high demand to automatically determine the optimal subset. In this study, we conduct pioneering work to explore the decision variant after obtaining a list of candidate subsets from RFE. We provide a detailed analysis and comparison of several decision variants to automatically select the optimal feature subset. Random forest (RF)-recursive feature elimination (RF-RFE) algorithm and a voting strategy are introduced. We validated the variants on two totally different molecular biology datasets, one for a toxicogenomic study and the other one for protein sequence analysis. The study provides an automated way to determine the optimal feature subset when using RF-RFE.