Genetic algorithms combined with discriminant analysis for key variable identification

High-order Fisher's discriminant analysis

A new key variable selection and prediction model of IAQ that can select key variables governing indoor air quality (IAQ), such as PM10, CO2, CO, VOCs and formaldehyde, are suggested in this paper. The essential problem of the prediction model is the question of which of the original variables are the most important for predicting IAQ. The next issue is determining the number of key variables that should be ranked. A new index of discriminant importance in the projection (DIP) of Fisher’s linear discriminant (FLD) is suggested for selecting key variables of the prediction models with multiple linear regression (MLR) and partial least squares (PLS), as well as for ranking the importance of input measurement variables on IAQ prediction. The prediction models were applied to a real IAQ dataset from telemonitoring data (TMS) in a metro system. The prediction results of the model using all variables were compared with the results of the model using only key variables of DIP. It shows that the use of our new variable selection method cannot only reduce computational effort, but will also enhance the prediction performances of the models.

This paper characterizes an approach to data compression and defect isolation for the multi-variable process by introducing the wavelet principal component analysis (PCA) and the genetic algorithms. In the defect analysis process, data compression is a critical phase. In our case of study, this can be achieved through the wavelet PCA. We can, through the use of such a technique, filter the data noise and the measurement errors by combining the PCA and the wavelet analysis. The genetic algorithm optimization feature is then employed in structuring the residues of the defect isolation. We present the principle of the PCA as well as the wavelet compression property. Then, we describe both the defect isolation classical method by structuring the residues and the principle of optimizing a problem by the genetic algorithms. The proposed approach is implemented on a statistic system and the Tennessee Eastman process. In comparison with other previous methods, the obtained results mirror the performance of the approach to defect isolation.

Novel online discriminant analysis based schemes to deal with observations from known and new classes: Application to industrial systems

Robustness of multivariate image analysis assessed by resampling techniques and applied to FDG-PET scans of patients with Alzheimer's disease

Interpretability of selected variables and performance comparison of variable selection methods in a polyethylene and polypropylene NIR classification task

In this study, a large number of features that were obtained to classify speech emotions were projected into different spaces, selecting different numbers of principal components in principal component analysis and Fisher's discriminant analysis. Classifications were performed in those spaces using Naive-Bayes classifier and obtained results were compared. While the highest accuracy obtained in the Fisher space was 57.87%, it was calculated as 48.02% in the principal component space.

This study illustrates a novel model based FDI method for the common mechanical faults arising during the manufacture of loudspeakers. To overcome the drawbacks of the conventional signal based approaches, the Bayesian classification of impulse responses based on a model based fault symptom database is proposed. The loudspeaker model is estimated via IRES and ARMA techniques. The fault symptom database is constructed with a novel nonlinear loudspeaker model. The performances of Principal Component Analysis (PCA) and Fisher’s Discriminant Analysis (FDA) are compared. The results show the effectiveness of the proposed method. It is also shown that the FDA based classifier performs better than PCA in terms of the accuracy and consistency of the healthy baseline estimation. However, the fault isolation is difficult due to the similarities of fault signatures.

With the development of data mining technology, it becomes vital to extract fault features effectively in fault diagnosis. This paper is mainly concerned about a hybrid method for fault diagnosis. Firstly, train the data with the deep belief networks (DBN) to get the weighting vectors for all the fault class. And then, weight for fault data with responding weight vector to obtain the summed data. Finally, use Fisher Discriminant Analysis (FDA) to do on the summed data sets, in order to diagnose the fault. Experimental results show the effective performance of this method with the Tennessee Eastman Process (TEP) benchmark.

In this paper, we describe the use of various methods of one-dimensional spectral compression by variable selection as well as principal component analysis (PCA) for compressing multi-dimensional sets of spectral data. We have examined methods of variable selection such as wavelength spacing, spectral derivatives, and spectral integration error. After variable selection, reduced transmission spectra must be decompressed for use. Here we examine various methods of interpolation, e.g., linear, cubic spline and piecewise cubic Hermite interpolating polynomial (PCHIP) to recover the spectra prior to estimating at-sensor radiance. Finally, we compressed multi-dimensional sets of spectral transmittance data from moderate resolution atmospheric transmission (MODTRAN) data using PCA. PCA seeks to find a set of basis spectra (vectors) that model the variance of a data matrix in a linear additive sense. Although MODTRAN data are intricate and are used in nonlinear modeling, their base spectra can be reasonably modeled using PCA yielding excellent results in terms of spectral reconstruction and estimation of at-sensor radiance. The major finding of this work is that PCA can be implemented to compress MODTRAN data with great effect, reducing file size, access time and computational burden while producing high-quality transmission spectra for a given set of input conditions.

Summary form only given. In the paper, the comparison of discriminative capabilities of conventional dimensionality reduction methods and the integration of a dimensionality reduction module and recognizer design with minimum classification error rate are discussed. Conventionally, principal component analysis (PCA) and Fisher's linear discriminant (FLD) are two most popular and widely used dimensionality reduction methods for pattern recognition. However, the objectives of these two methods are quite different. PCA basically tries to faithfully keep the original data representation but FLD tries to separate data distribution of different classes. It therefore seems that FLD can provide more discriminative characteristics to pattern recognition than PCA. However, a completely optimal feature extractor can never be anything but an optimal recognizer. It is only when constraints are placed on the classifier that one can formulate nontrivial problems. We apply a minimum error formulation (MEF) to integrate the design of dimensionality reduction module and pattern recognizer. The experimental results show that such an integration provides very good recognition performance on a data set of hand-written Chinese characters even when the feature number has been significantly reduced.

An electronic nose (E-nose) was developed to distinguish different stored eggs by using principal component analysis (PCA),linear discriminant analysis (LDA),BP neural network (BPNN) and the combination of a genetic algorithm and BP neural network (GANN). Results showed that the E-nose can distinguish eggs of different storage time under cool and room-temperature storage by LDA and PCA. LDA preceded PCA. Better prediction values were obtained by GANN than by BPNN. Relationships were established between the E-nose signal and egg quality indices by quadratic polynomial step regression (QPSR). The prediction models indicated good prediction performance with correlation coefficient between predicted and measured values greater than 0.90.

This study focuses on the fault detection (FD) and false alarm rates (FAR) of Principal component analysis (PCA) and independent component analysis (ICA) algorithms on the Tennessee Eastman (TE) process. However, PCA and ICA algorithms have been applied widely to systems for data driven fault detection, there are limited work on FARs of the algorithms. In this work, FARs of the algorithms are investigated on TE process. Simulation study indicates that the proposed algorithms are robust for fault detection, and ICA has higher performance than PCA for FARs.

The essential task of a Brain-Computer Interface (BCI) is to extract the motor imagery features from Electro-Encephalogram (EEG) signals for classifying the thought process. It is necessary to analyse these obtained signals in both the time domain and frequency domains. It is observed that the combination of multiple algorithms increases the performance of the feature extraction process. This paper identifies combinations that have not been attempted previously and improves the accuracy of the overall process, although other authors implemented different combinations of the techniques. The focus is given more on the feature extraction process and frequency bands, which are user-specific and subject-specific frequency bands. In both time and frequency domains, after analysing EEG signals with the time domain parameter, we select the frequency band and the timing while using the Fisher ratio of the time domain parameter (TDP). We used Fisher discriminant analysis (FDA)-type F-score to simultaneously select the frequency band and time segment for multi-class classification. We extracted subject-specific TDP features from the training trials to train the classifier when optimal time-frequency areas were selected for each subject. In this paper, various methods are explored for obtaining the features, which are Time Domain Parameters (TDP), Fast Fourier Transform (FFT), Principal Component Analysis (PCA), R2, Fast Correlation Based Filter (FCBF), Empirical Mode Decomposition (EMD), and Intrinsic time-scale decomposition (ITD). After the extraction process, PCA is used for dimensionality reduction. An efficient result was obtained with the combination of TDP, FFT, and PCA. We used the multi-class Fisher’s linear discriminant analysis (LDA) as the classifier, which was in line with the FDA-type F-score. It is observed that the combination of feature extraction techniques to the frequency bands that were selected by the Fisher ratio and FDA type F-score along with Fisher’s LDA classifier had higher accuracy than the results obtained other researches. A kappa coefficient accuracy of 0.64 is obtained for the proposed technique. Our method leads to better classification performance when compared to state-of-the-art methods. The novelty of the approach is based on the combination of frequency bands and two feature extraction methods.

Rapid and accurate identification of foodborne pathogenic bacteria is of great importance because they are often responsible for the majority of serious foodborne illnesses. The confocal Raman microspectroscopy (CRM) is a fast and easy-to-use method known for its effectiveness in detecting and identifying microorganisms. This study demonstrates that CRM combined with chemometrics can serve as a rapid, reliable, and efficient method for the detection and identification of foodborne pathogenic bacteria without any laborious pre-treatments. Six important foodborne pathogenic bacteria including S. flexneri, L. monocytogenes, V. cholerae, S. aureus, S. typhimurium, and C. botulinum were investigated with CRM. These pathogenic bacteria can be differentiated based on several characteristic peaks and peak intensity ratio. Principal component analysis (PCA) was used for investigating the difference of various samples and reducing the dimensionality of the dataset. Performances of some classical classifiers were compared for bacterial detection and identification including decision tree (DT), artificial neural network (ANN), and Fisher's discriminant analysis (FDA). Correct recognition ratio (CRR), area under the receiver operating characteristic curve (ROC), cumulative gains, and lift charts were used to evaluate the performance of models. The impact of different pretreatment methods on the models was explored, and pretreatment methods include Savitzky-Golay algorithm smoothing (SG), standard normal variate (SNV), multivariate scatter correction (MSC), and Savitzky-Golay algorithm 1st Derivative (SG 1st Der). In the DT, ANN, and FDA model, FDA is more robust for overfitting problem and offers the highest accuracy. Most pretreatment methods raised the performance of the models except SNV. The results revealed that CRM coupled with chemometrics offers a powerful tool for the discrimination of foodborne pathogenic bacteria.