Incorporating linear discriminant analysis in neural tree for multidimensional splitting

In this paper, a new hybrid classifier is proposed by combining neural network and direct fractional-linear discriminant analysis (DF-LDA). The proposed hybrid classifier, neural tree with linear discriminant analysis called NTLD, adopts a tree structure containing either a simple perceptron or a linear discriminant at each node. The weakly performing perceptron nodes are replaced with DF-LDA in an automatic way. Taking the advantage of this node substitution, the tree building process converges faster and avoids the over-fitting of complex training sets in training process resulting a shallower tree together with better classification performance. The proposed NTLD algorithm is tested on various synthetic and real datasets. The experimental results show that the proposed NTLD leads to very satisfactory results in terms of tree depth reduction as well as classification accuracy.     

Artificial neural networks for optimization of gold-bearing slime smelting

Pyrometallurgy is often used in the industrial process for treating gold-bearing slime. Slag compositions have remarkable influences on the recovery of gold and the gold content in slag. A method for determining optimum flux compounding with neural networks is studied in this paper, and the neural network model for estimating the gold contents with different slag compositions is presented. On the basis of the neural network model, an algorithm for searching the optimum flux compounding in the gold-slime smelting process is proposed, and the optimum flux compositions are obtained accordingly.     

A new local search based hybrid genetic algorithm for feature selection

This paper presents a new hybrid genetic algorithm (HGA) for feature selection (FS), called as HGAFS. The vital aspect of this algorithm is the selection of salient feature subset within a reduced size. HGAFS incorporates a new local search operation that is devised and embedded in HGA to fine-tune the search in FS process. The local search technique works on basis of the distinct and informative nature of input features that is computed by their correlation information. The aim is to guide the search process so that the newly generated offsprings can be adjusted by the less correlated (distinct) features consisting of general and special characteristics of a given dataset. Thus, the proposed HGAFS receives the reduced redundancy of information among the selected features. On the other hand, HGAFS emphasizes on selecting a subset of salient features with reduced number using a subset size determination scheme. We have tested our HGAFS on 11 real-world classification datasets having dimensions varying from 8 to 7129. The performances of HGAFS have been compared with the results of other existing ten well-known FS algorithms. It is found that, HGAFS produces consistently better performances on selecting the subsets of salient features with resulting better classification accuracies.     

监测挥发性有机物的SAW传感器的研究进展

综述了可以连续操作和用于原位监测有机挥发性化合物的声表面波(SAW)化学传感器的研究进展。不锈钢包装和集成电路板的应用,使SAW传感器可以用于空气、土壤和水等在内的多种环境中监测有机挥发性化合物(VOCs)。介绍了根据常见地下水污染物氯代脂肪烃(如三氯乙烯)选择的高分子膜以及对传感器进行的优化。     

垃圾图像判别中的特征提取与选择研究*

对垃圾图像判别问题中的特征提取和特征选择研究现状进行了总结。从特征的可区分性、鲁棒性和提取效率三个方面比较了垃圾图像判别中的主要特征,分析了特征的优缺点。结合分类学习算法、仿真实验结果,对已有的主要特征选择和分析方法进行比对,为进一步研究特征提取、特征选择方法,提高垃圾图像分类器的性能和效率提供有价值的参考。     

Adaptive resonance neural classifier for identification of gases/odours using an integrated sensor array

A new approach to intelligent gas sensor (IGS) design using a classifier based on adaptive resonance theory (ART) artificial neural network (ANN) is presented. Using published data of sensor arrays fabricated and characterised at our laboratory, we demonstrate excellent gas/odour identification performance of our classifier for a 4-gas, 4-sensor system to identify individual gas/odour. Since the ART neural network is a self-organising classifier trained in the unsupervised mode, it avoids the drawbacks associated with static feedforward neural networks trained with a locally optimal backpropagation-type training algorithms applied by researchers in the recent past. The ART neural network offers easy implementability and real time performance in addition to giving excellent classification accuracy as demonstrated by our experiments.     

GA-based learning bias selection mechanism for real-time scheduling systems

The use of machine learning technologies in order to develop knowledge bases (KBs) for real-time scheduling (RTS) problems has produced encouraging results in recent researches. However, few researches focus on the manner of selecting proper learning biases in the early developing stage of the RTS system to enhance the generalization ability of the resulting KBs. The selected learning bias usually assumes a set of proper system features that are known in advance. Moreover, the machine learning algorithm for developing scheduling KBs is predetermined. The purpose of this study is to develop a genetic algorithm (GA)-based learning bias selection mechanism to determine an appropriate learning bias that includes the machine learning algorithm, feature subset, and learning parameters. Three machine learning algorithms are considered: the back propagation neural network (BPNN), C4.5 decision tree (DT) learning, and support vector machines (SVMs). The proposed GA-based learning bias selection mechanism can search the best machine learning algorithm and simultaneously determine the optimal subset of features and the learning parameters used to build the RTS system KBs. In terms of the accuracy of prediction of unseen data under various performance criteria, it also offers better generalization ability as compared to the case where the learning bias selection mechanism is not used. Furthermore, the proposed approach to build RTS system KBs can improve the system performance as compared to other classifier KBs under various performance criteria over a long period.     

Comparison of variable selection methods for PLS-based soft sensor modeling

Data-driven soft sensors have been widely used in both academic research and industrial applications for predicting hard-to-measure variables or replacing physical sensors to reduce cost. It has been shown that the performance of these data-driven soft sensors could be greatly improved by selecting only the vital variables that strongly affect the primary variables, rather than using all the available process variables. In this work, a comprehensive evaluation of different variable selection methods for PLS-based soft sensor development is presented, and a new metric is proposed to assess the performance of different variable selection methods. The following seven variable selection methods are compared: stepwise regression (SR), partial least squares with regression coefficients (PLS-BETA), PLS with variable importance in projection (PLS-VIP), uninformative variable elimination with PLS (UVE-PLS), genetic algorithm with PLS (GA-PLS), least absolute shrinkage and selection operator (Lasso), and competitive adaptive reweighted sampling with PLS (CARS-PLS). Their strengths and limitations for soft sensor development are demonstrated by a simulated case study and an industrial case study.     

Analysis of new variable selection methods for discriminant analysis

Several methods to select variables that are subsequently used in discriminant analysis are proposed and analysed. The aim is to find from among a set of m variables a smaller subset which enables an efficient classification of cases. Reducing dimensionality has some advantages such as reducing the costs of data acquisition, better understanding of the final classification model, and an increase in the efficiency and efficacy of the model itself. The specific problem consists in finding, for a small integer value of p, the size p subset of original variables that yields the greatest percentage of hits in the discriminant analysis. To solve this problem a series of techniques based on metaheuristic strategies is proposed. After performing some test it is found that they obtain significantly better results than the stepwise, backward or forward methods used by classic statistical packages. The way these methods work is illustrated with several examples.     

Evaluation of feature selection methods for text classification with small datasets using multiple criteria decision-making methods

The evaluation of feature selection methods for text classification with small sample datasets must consider classification performance, stability, and efficiency. It is, thus, a multiple criteria decision-making (MCDM) problem. Yet there has been few research in feature selection evaluation using MCDM methods which considering multiple criteria. Therefore, we use MCDM-based methods for evaluating feature selection methods for text classification with small sample datasets. An experimental study is designed to compare five MCDM methods to validate the proposed approach with 10 feature selection methods, nine evaluation measures for binary classification, seven evaluation measures for multi-class classification, and three classifiers with 10 small datasets. Based on the ranked results of the five MCDM methods, we make recommendations concerning feature selection methods. The results demonstrate the effectiveness of the used MCDM-based method in evaluating feature selection methods.     

A new feature selection method based on association rules for diagnosis of erythemato-squamous diseases

In this paper, a new feature selection method based on Association Rules (AR) and Neural Network (NN) is presented for the diagnosis of erythemato-squamous diseases. AR is used for reducing the dimension of erythemato-squamous diseases dataset and NN is used for efficient classification. The proposed AR+NN system performance is compared with that of other feature selection algorithms+NN. The dimension of input feature space is reduced from thirty four to twenty four by using AR. In test stage, 3-fold cross validation method is applied to the erythemato-squamous diseases dataset to evaluate the proposed system performances. The correct classification rate of proposed system is 98.61%. This research demonstrated that the AR can be used for reducing the dimension of feature space and proposed AR+NN model can be used to obtain fast automatic diagnostic systems for other diseases.     

面向宏观地表分类的特征选择算法比较研究

为了成功将土地覆盖进行分类,选择合适的特征是至关重要的。针对利用MODIS数据进行宏观土地覆盖的分类问题,对三种典型的特征选择方法进行了比较研究。研究结果表明：分支定界法（BB）最适合于该土地覆盖分类问题,与此同时,ReliefF和mRMR方法在目标应用中的精度非常接近。研究结果同样表明进行特征选择是非常必要的,它不仅能够大大地降低计算复杂度,而且分类精度能够保持不变,甚至更高。     

Genetic algorithms for feature selection of image analysis-based quality monitoring model: An application to an iron mine

Measuring the quality parameters of materials at mines is difficult and a costly job. In this paper, an image analysis-based method is proposed efficiently and cost effectively that determines the quality parameters of material. The image features are extracted from the samples collected from a mine and modeled using neural networks against the actual grade values of the samples generated by chemical analysis. The dimensions of the image features are reduced by applying the genetic algorithm. The results showed that only 39 features out of 189 features are sufficient to model the quality parameter. The model was tested with the testing data set and the result revealed that the estimated grade values are in good agreement with the real grade values (R²=0.77). The developed method was then applied to a case study mine of iron ore. The case study results show that proposed image-based algorithm can be a good alternative for estimating quality parameters of materials at a mine site. The effectiveness of the proposed method was verified by applying it on a limestone deposit and the results revealed that the method performed equally well for the limestone deposit.     

Development of a variable selection method for soft sensor using artificial neural network and nonnegative garrote

This paper developed a new variable selection method for soft sensor applications using the nonnegative garrote (NNG) and artificial neural network (ANN). The proposed method employs the ANN to generate a well-trained network, and then uses the NNG to conduct the accurate shrinkage of input weights of the ANN. This paper took Bayesian information criterion as the model evaluation criterion, and the optimal garrote parameter s was determined by v-fold cross-validation. The performance of the proposed algorithm was compared to existing state-of-art variable selection methods. Two artificial dataset examples and a real industrial application for air separation process were applied to demonstrate the performance of the methods. The experimental results showed that the proposed method presented better model accuracy with fewer variables selected, compared to other state-of-art methods.     

B-spline neural network design using improved differential evolution for identification of an experimental nonlinear process

B-Spline Neural Network (BSNN), a type of basis function neural network, is trained by gradient-based methods which may fall into local minima during the learning procedure. To overcome the limitations encountered by gradient-based optimization methods, we propose differential evolution (DE) – an evolutionary computation methodology – which can provide a stochastic search to adjust the control points of a BSNN. In this paper, we propose six DE approaches using chaotic sequences based on logistic mapping to train a BSNN. Chaos describes the complex behavior of a nonlinear deterministic system. The application of chaotic sequences instead of random sequences in DE is a powerful strategy to diversify the DE population and improve the DE's performance in preventing premature convergence to local minima. The numerical results presented here indicate that chaotic DE was effective for building a good BSNN model for the nonlinear identification of an experimental nonlinear yo–yo motion control system.     

A fully neural implementation of unitary response model for classification of gases/odors using the responses of thick film gas sensor array

In this paper, a classification scheme based on neurally implemented unitary response model (URM) for a gas/odor sensor array response has been presented. Thick-film tin-oxide sensor array responses for four gases/odors (viz. acetone, carbon tetra-chloride, ethyl methyl ketone and xylene) were first transformed into equivalent unitary responses. This transformation was carried out using a pre-trained neural ‘unitary response model pre-processor (URMP)’, called Net I_URMP. The classification of these responses in the unitary analysis space was then carried out, more accurately, using a pre-trained neural classifier called Net II_URMC. During this experiment, respective nets Net I_URMP and Net II_URMC, comprising of 12 and 8 neurons, were trained in just 23 and 09 epochs of 42 × 4 training response vectors. At stage I, the mean squared error (MSE) between neurally and mathematically obtained unitary response versions of 18 independent test responses for the considered gases/odors was 7.51 × 10⁻². At stage II, all the aforesaid test samples were correctly classified, with a MSE of 3.87 × 10⁻⁸. Further, by connecting Net I_URMP and Net II_URMC in cascade, the proposed classifier could be implemented using 16 neurons only.     

Taking advantage of improved resource allocating network and latent semantic feature selection approach for automated text categorization

In this study we propose an improved learning algorithm based on resource allocating network (RAN) for text categorization. RAN is a promising neural network of single hidden layer structure based on radial basis function. We firstly use the means clustering-based method to determine the initial centers in the hidden layer. Such method can effectively overcome the limitation of local-optimal of clustering algorithms. Subsequently, in order to improve the novelty criteria of RAN, we propose a root mean square (RMS) sliding window method which can reduce the underlying influence of undesirable noise data. Through the further research on the learning process of RAN, we divide the learning process of RAN into a preliminary study phase and a subsequent study phase. The former phase initializes the preliminary structure of RAN and decreases the complexity of network, while the latter phase refines its learning ability and improves the classification accuracy. Such a compact network structure decreases the computational complexity and maintains the higher convergence rate. Moreover, a latent semantic feature selection method is utilized to organize documents. This method reduces the input scale of network, and reveals the latent semantics between features. Extensive experiments are conducted on two benchmark datasets, and the results demonstrate the superiority of our algorithm in comparison with state of the art text categorization algorithms.     

Novel e-nose for the discrimination of volatile organic biomarkers with an array of carbon nanotubes (CNT) conductive polymer nanocomposites (CPC) sensors

We report the original design of a new type of electronic nose (e-nose) consisting of only five sensors made of hierarchically structured conductive polymer nanocomposites (CPC). Each sensor benefits from both the exceptional electrical properties of carbon nanotubes (CNT) used to build the conductive architecture and the spray layer by layer (sLbL) assembly technique, which provides the transducers with a highly specific 3D surface structure. Excellent sensitivity and selectivity were obtained by optimizing the amount of CNT with five different polymer matrices: poly(caprolactone) (PCL), poly(lactic acid) (PLA), poly(carbonate) (PC), poly(methyl methacrylate) (PMMA) and a biobased polyester (BPR). The ability of the resulting e-nose to detect nine organic solvent vapours (isopropanol, tetrahydrofuran, dichloromethane, n-heptane, cyclohexane, methanol, ethanol, water and toluene), as well as biomarkers for lung cancer detection in breath analysis, has been demonstrated. Principal component analysis (PCA) proved to be an excellent pattern recognition tool to separate vapour clusters.     

Use of high-resolution satellite imagery in an integrated model to predict the distribution of shade coffee tree hybrid zones

In New Caledonia (21°S, 165°E), shade-grown coffee plantations were abandoned for economic reasons in the middle of the 20th century. Coffee species (Coffea arabica, C. canephora and C. liberica) were introduced from Africa in the late 19th century, they survived in the wild and spontaneously cross-hybridized. Coffee species were originally planted in native forest in association with leguminous trees (mostly introduced species) to improve their growth. Thus the canopy cover over rustic shade coffee plantations is heterogeneous with a majority of large crowns, attributed to leguminous trees. The aim of this study was to identify suitable areas for coffee inter-specific hybridization in New Caledonia using field based environmental parameters and remotely sensed predictors. Due to the complex structure of tropical vegetation, remote sensing imagery needs to be spatially accurate and to have the appropriate bands for monitoring vegetation cover. Quickbird panchromatic (black and white) imagery at 0.6 to 0.7 m spatial resolutions and multispectral imagery at 2.4 m spatial resolution were pansharpened and used for this study. The two most suitable remotely sensed indicators, canopy heterogeneity and tree crown size, were acquired by the sequential use of tree crown detection (neural network), image processing (such as textural analysis) and classification. All models were supervised and trained on learning data determined by human expertise. The final model has two remotely sensed indicators and three physical parameters based on the Digital Elevation Model: elevation, slope and water flow accumulation. Using these five predictive variables as inputs, two modelling methods, a decision tree and a neural network, were implemented. The decision tree, which showed 96.9% accuracy on the test set, revealed the involvement of ecological parameters in the hybridization of Coffea species. We showed that hybrid zones could be characterized by combinations of modalities, underlining the complexity of the environment concerned. For instance, forest heterogeneity and large crown size, steep slopes (> 53.5%) and elevation between 194 and 429 m asl, are favourable factors for Coffea inter-specific hybridization. The application of the neural network on the whole area gave a predictive map that distinguished the most suitable areas by means of a nonlinear continuous indicator. The map provides a confidence level for each area. The most favourable areas were geographically localized, providing a clue for the detection and conservation of favourable areas for Coffea species neo-diversity.     

Comparison of Machine Learning Methods for Land Use/Land Cover Classification in the Complicated Terrain Regions

Aiming at the characteristics of varied and complex geomorphic types,crisscross network of ravines and broken terrain in high altitude complicated terrain regions,it is very important to study and find the rapid and effective land use/land cover classification method for obtaining and timely updating of land use information.Taking the Huangshui river basin located in the transitional zone between the Loess Plateau and the Qinghai-Tibet Plateau as acasestudy area,the objective of this study is to explore a kind of effective information extraction method from comparison of four kinds machine learning methods for complicated terrain regions.based on Landsat 8 OLI satellite data,DEM and combined with various thematic features,on the basis of geographical division of the study area,artificial neural network,decision tree,support vector machine and random forest four machine learning methods for land use information extraction were used to obtain land use data,and confusion matrix was constructed to evaluate classification accuracy.The results showed that the classification accuracies of random forest and decision tree are obviously higher than those of support vector machine and artificial neural network.The random forest method has the highest classification accuracy,the overall classification accuracy is 85.65%,the Kappa coefficient is 0.84.based on the above classification,Random forest classification method was chose to further classify Landsat 8 fusion datafrom panchromatic 15 meter and multispectral 30 meter image,the overall classification accuracy is 86.49% and the Kappa coefficient is 0.85.This indicated that the random forest classification method can obtain higher classification efficiency while ensuring the classification accuracy.It is very effective for the extraction of land use information in complicated terrain regions.Data fusion can improve the classification accuracy to a certain extent.