Non-Negative Matrix Factorization-Based EEG Signal Classification

Non-negative matrix factorization (NMF) has been one of the most popular methods for feature learning in the field of machine learning and computer vision. Most existing works directly apply NMF on high-dimensional image datasets for computing the effective representation of the raw images. However, in fact, the common essential information of a given class of images is hidden in their low rank parts. For obtaining an effective low-rank data representation, we in this paper propose a non-negative low-rank matrix factorization (NLMF) method for image clustering. For the purpose of improving its robustness for the data in a manifold structure, we further propose a graph regularized NLMF by incorporating the manifold structure information into our proposed objective function. Finally, we develop an efficient alternating iterative algorithm to learn the low-dimensional representation of low-rank parts of images for clustering. Alternatively, we also incorporate robust principal component analysis into our proposed scheme. Experimental results on four image datasets reveal that our proposed methods outperform four representative methods.

The disruption has been occurring in financial services. Rethinking new strategy of banking is needed to make a sustainable competitive advantage innovation in organizations. The four types of business strategy for banks are prospector, analyzer, defender, and reactor. Studies mentioned that formulating strategy is very costly, time consuming, and comprehensive analysis. The banks have to get rid of execution time inefficiency, lack of flexibility, and lack of ability to present several scenarios in the dynamic business environment. The purpose of this study is to present an integrated intelligence algorithm for estimating strategic resources of the bank strategy in Indonesia. The algorithm has two basic modules which are artificial neural network (ANN) and analytical hierarchy process (AHP). ANN is utilized as an inductive algorithm in discovering predictive strategy of the bank and used to explain the strategic resources which improved in forward. AHP has the capability to handle multi-level decision-making structure with use of expert judgments in pairwise comparison process. AHP is used to measure the weight of the resources and the score is used to determine the strategy. The empirical results indicate that ANN and AHP integration is proved to predict the business strategy of the bank. The strategy choice appropriate with the condition of bank's resources. This framework can be implemented to help banker for the decision making in bank operation. Keywords: business strategy, ANN, AHP, resources

Concrete compressive strength is recognized as one of the most important mechanical properties of concrete and one of the most significant mechanical properties in determining the quality of the produced concrete. Since the traditional procedures of determining the compressive strength of concrete require time and cost, scholars have always been looking for new methods to replace them with existing traditional methods. In this paper, soft computing methods are investigated for determining the compressive strength of concrete. To be specific, 150 different concrete specimens with various mix design parameters have been built in the laboratory, and the compressive strength of them have been measured after 28 days of curing in the water. Five different concrete mix parameters, (i.e., cement, water to cement ratio, gravel, sand, and microsilica) were considered as input variables. In addition, two soft computing techniques have been used in this study which are Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference (ANFIS) System. Results have shown that both of ANN and ANFIS models are successful models for predicting the compressive strength of concrete. Also, results have shown that ANFIS is more capable than ANN in predicting the compressive strength of concrete.

In this work, we will consider the dimension reduction of the set of time series, such as economic data, to find the meaningful basis vector for the set of data, and indicate which data use which basis vector. Usually each of the time series is analyzed independently in economics but here we will analyze the set of time series simultaneously. Since some of the economic data are measured as positive values and we want to decompose them as a mixture of the parts, we will apply non-negative matrix factorization to the economic data. Non-negative matrix factorization can compress dimensions by approximating a non-negative matrix with the product of two non-negative matrices. The two non-negative matrices are called the coefficient matrix and the basis matrix, and the basis matrix can be considered as a dimensionally compressed matrix. If the standard non-negative matrix factorization is used for economic data, the basis matrix may not be smooth. We think that the basis vectors should be smooth except a few special economical incidents. In the proposed method, a Graph-based non-negative matrix factorization is introduced to regularize the basis matrix of the time series. A path graph for representing the time series of economic data is incorporated into the non-negative matrix factorization as regularization. As a result, basis vectors that maintains the time series of economic data are decomposed. Furthermore, we propose to introduce a sparsity in the non-negative matrix factorization. Traditionally, the sparsity incorporated into non-negative matrix factorization has been used for basis vectors. However, the proposed method introduces the sparsity for coefficient vectors. Thus the proposed method, which simultaneously incorporates the sparsity for the coefficient vectors and the smoothness for the basis vectors, can extract the smooth basis vectors and the original economical data are approximated as the weighted sum of the few bases vectors. This allows us to discover economic trends and the best-fit trends for each data at the same time.

The primary objective of Short-Term Load Forecasting, often known as STLF, is to forecast load with a lead time of anything from one hour to seven days. In this study, we suggest the use of AI approaches for short-term load forecasting. Some examples of these techniques are artificial neural networks (ANN) and wavelet neural networks (WNN) (STLF). A reliable prediction makes the challenge of managing generation and load much more manageable. The ANN and WNN algorithms are used in order to calculate the STLF for the next day. The comparison between the two approaches as well as their performance is outlined, and the normalized MAPE error for one day ahead is shown in the article as well. In addition to this, the Validation is performed on the TNEB testing system. This study provided an application of AI techniques, specifically ANN and WNN, for STLF applications in power systems. For hourly load forecasting, analyses of the capabilities and characteristics of ANN and WNN are done. The suggested strategies have already been tried and tested in actual world settings with great success.

Plant diseases and insect pests have similar symptoms, but it is difficult to distinguish between professional and technical personnel to identify plant diseases and insect pests.In order to accurate extraction of plant diseases and insect pests, physiological and pathological characteristics of signal, puts forward a based on lifting wavelet transform feature extraction algorithm optimization scheme, for the study of plant diseases and insect pests damage signal showing the effect of the prior farmers identify any disease, choose the correct method of governance, quickly make the right decision, improve farmers plant diseases and insect pests, harm signal feature extraction and recognition level.The simulation results show that this algorithm can be used to optimize the stability and convergence, and can be used as an ideal plant disease and insect pests signal feature extraction optimization algorithm, which can effectively identify the different plant diseases and insect pests.

Profile-Decomposing Output of Multi-Channel Odor Sensor Array

Nonnegative matrix factorization (NMF) is the problem of approximating an input nonnegative matrix, $V$, as the product of two smaller nonnegative matrices, $W$ and $H$. In this paper, we introduce a general framework to design multiplicative updates (MU) for NMF based on $\beta$-divergences ($\beta$-NMF) with disjoint equality constraints, and with penalty terms in the objective function. By disjoint, we mean that each variable appears in at most one equality constraint. Our MU satisfy the set of constraints after each update of the variables during the optimization process, while guaranteeing that the objective function decreases monotonically. We showcase this framework on three NMF models, and show that it competes favorably the state of the art: (1)~$\beta$-NMF with sum-to-one constraints on the columns of $H$, (2) minimum-volume $\beta$-NMF with sum-to-one constraints on the columns of $W$, and (3) sparse $\beta$-NMF with $\ell_2$-norm constraints on the columns of $W$.

Formation pressure is a critical formation condition that affects the efficiency and the economy of drilling operations. The knowledge of the formation pressure is significant to control the well. It will assist in avoiding problems associated with drilling operation and decreasing the cost of the drilling operation. It is essential to predict formation pressure accurately prior to the drilling process to prevent circulation loss and kick. Many methods are used to estimate the formation pressure either from log information or drilling parameters. These methods need pressure trends such as abnormal or normal to estimate the formation pressure. Only two papers used artificial intelligence (AI) to estimate the formation pressure by applying only two AI techniques. In this paper, actual field measurements such as log data (formation density (RHOB), porosity (ϕ), and compressional time (Δt)) and drilling parameters (weight on bit (WOB), rotary speed (RPM), penetration rate (ROP), mud density (MW)) were utilized by five techniques of AI to predict the pore pressure. These AI techniques are artificial neural networks (ANN), radial basis function (RBF), fuzzy logic (FL), support vector machine (SVM), and functional networks (FN). In addition, comparative analyses were performed to determine the best combination of AI technique to predict the pore pressure with high accuracy. All artificial intelligence techniques predicted the formation pressure with high performance. In addition, the developed AI methods can be used to estimate the formation pressure without the need for normal pressure trends. The comparative analysis results show that support vector machine has the feature of formation pressure estimation by its short time prediction and high accuracy (a coefficient of determination of 0.995 and an average percentage error of 0.14%). It is recommended to apply SVM to predict the pore pressure in field operations.

A novel approach to facial expression recognition (FER) based on the combination of Non-negative Matrix Factorization (NMF) and Support Vector Machine (SVM) was proposed. One key step in FER is to extract expression features from the original face images. NMF is an effective approach to extract expression features because NMF decomposition makes the reconstruction of expression images in a non-subtractive way and is much similar to the process of forming unity from parts. The proposed algorithm first processes facial expression image with histogram equalization operator. Then NMF method is used for feature dimension reduction and SVM for classification. Finally, the algorithm was implemented with Matlab and experimented in Japanese female facial expression database (JAFEE database). A recognition rate of 66.19% was obtained and showed the effectiveness of the proposed algorithm

High-resolution liquid chromatography-tandem mass spectrometry (LC--MS-MS)-based machine learning models are constructed to address the analytical challenge of identifying unknown controlled substances and new psychoactive substances (NPSs). Using a training set composed of 770 LC-MS-MS barcode spectra (with binary entries 0 or 1) obtained generally by high-resolution mass spectrometers, three classification machine learning models were generated and evaluated. The three models are artificial neural network (ANN), support vector machine (SVM) and k-nearest neighbor (k-NN) models. In these models, controlled substances and NPSs were classified into 13 subgroups (benzylpiperazine, opiate, benzodiazepine, amphetamine, cocaine, methcathinone, classical cannabinoid, fentanyl, 2C series, indazole carbonyl compound, indole carbonyl compound, phencyclidine and others). Using 193 LC-MS-MS barcode spectra as an external test set, accuracy of the ANN, SVM and k-NN models were evaluated as 72.5%, 90.0% and 94.3%, respectively. Also, the hybrid similarity search (HSS) algorithm was evaluated to examine whether this algorithm can successfully identify unknown controlled substances and NPSs whose data are unavailable in the database. When only 24 representative LC-MS-MS spectra of controlled substances and NPSs were selectively included in the database, it was found that HSS can successfully identify compounds with high reliability. The machine learning models and HSS algorithms are incorporated into our home-coded artificial intelligence screener for narcotic drugs and psychotropic substances standalone software that is equipped with a graphic user interface. The use of this software allows unknown controlled substances and NPSs to be identified in a convenient manner.

Repetitive transcranial magnetic stimulation (rTMS) is a non-pharmacological treatment that is associated with significant improvements in clinical symptoms of major depressive disorder (MDD). The efficacy of rTMS treatment can be predicted using pre-treatment cordance, a quantitative electroencephalography (QEEG) method extracting information from absolute and relative power of EEG spectra, that will prevent trial-and error treatment sequences, subject suffering and increase in health-care costs. In this study, pre-treatment QEEG data were collected from 6 frontal electrodes (Fp1, Fp2, F3, F4, F7 and F8) in slow bands (delta and theta) for 147 MDD subjects. In order to classify MDD subjects as responder or non-responder, four different machine learning techniques, which are Artificial Neural Networks (ANN), Support Vector Machines (SVM) and Decision Tree (DT), were used and their performances were evaluated. The results show that it is possible to predict rTMS treatment responders with a sensitivity of 95.6%, accuracy of 86.4% and area under Receiver Operating Characteristics (ROC) curve (AUC) value of 0.92 using SVM.

It is well understood that with a large number of data, an excellent interpretation of the subsurface condition can be produced, and also our understandings of the subsurface conditions can be improved significantly. However, having abundant subsurface geological and petrophysical data sometimes may not be possible, mainly due to budget issues. This situation can generate issues during hydrocarbon exploration and/or development activities.
 In this paper, the authors tried to apply artificial intelligence (AI) techniques to estimate outcomes values of particular wireline log data, using available petrophysic data. Two types of AI were selected and these are artificial neural network (ANN), and multiple linear regression (MLR). This research aims to advance our understanding of AI and its application in geology. There are three objectives of this study: (1) to estimate sonic log (DT) and density log (RhoB) using different types of AI (ANN and MLR); (2) to assess the best AI technique that can be used to estimate certain wireline log data; and (3) to compare the estimated wireline log values with the real, recorded values from the subsurface.
 Findings from this study show that ANN consistently provided a better accuracy percentage compared to MLR when estimating density log (RhoB). While using different set of data and technique, estimation of sonic log (DT) produced different accuracy level. Moreover, crossplot validation of the results show that the results from ANN analysis produced higher trendline reliability (R2) and correlation coefficient (R) than the results from MLR analysis. Comparison of the estimated RhoB and DT log data with the original recorded data shows minor mismatch. This is evident that AI technique can be a reliable solution to estimate particular outcomes of wireline log data, due to limited availability of the original recorded subsurface petrophysic data. It is expected that these findings would provide new insights into the application of AI in geology, and encourage the readers to explore and expand the many possibilities of the application of AI in geology.

Nearest neighbour classifier and support vector machine (SVM) are successful classifiers that are widely used in many important application areas. But both these classifiers suffer from the curse of dimensionality. Nearest neighbour search, in high dimensional data, using Euclidean distance is questionable since all the pair wise distances seem to be almost the same. In order to overcome this problem, we propose a novel classification system based on majority voting. Firstly, we partition the features into a number of blocks and construct a classifier for each block. The majority voting is then performed across all classifiers to determine the final class label. Classification is also performed using non-negative matrix factorisation (NNMF) that embeds high dimensional data into low dimensional space. Experiments were conducted on three of the benchmark datasets and the results obtained showed that the proposed system outperformed the conventional classification using both k-nearest neighbour (k-NN) and support vector machine (SVM) classifiers. The proposed system also showed better performance when compared with the classification performance of 1NN and SVM classifier using NNMF-based dimensionally reduced data.

RETRACTED: Detection of attacks in IoT sensors networks using machine learning algorithm