Test Error Research Articles

Tell Machine Learning Potentials What They Are Needed For: Simulation-Oriented Training Exemplified for Glycine.

Machine learning potentials (MLPs) are widely applied as an efficient alternative way to represent potential energy surfaces (PESs) in many chemical simulations. The MLPs are often evaluated with the root-mean-square errors on the test set drawn from the same distribution as the training data. Here, we systematically investigate the relationship between such test errors and the simulation accuracy with MLPs on an example of a full-dimensional, global PES for the glycine amino acid. Our results show that the errors in the test set do not unambiguously reflect the MLP performance in different simulation tasks, such as relative conformer energies, barriers, vibrational levels, and zero-point vibrational energies. We also offer an easily accessible solution for improving the MLP quality in a simulation-oriented manner, yielding the most precise relative conformer energies and barriers. This solution also passed the stringent test by diffusion Monte Carlo simulations.

Ultrasonic non-destructive inspection method for glass fibre reinforcement polymer (GFRP) hull plates considering design and construction characteristics

The design and construction characteristics, including glass fibre weight fraction ( Gc), number of single-ply layers, fabric combination, and fabrication quality, of glass fibre reinforcement polymer (GFRP) hull plates affect ultrasound propagation characteristics, such as ultrasonic velocity and attenuation, thereby influencing the accuracy of ultrasonic non-destructive test results. Therefore, this study is to propose a method to decrease the ultrasonic test errors of GFRP hull plate by using statistical method. The GFRP specimens with Gc of approximately 30–50 wt%, thicknesses of approximately 5–20 mm, and different fabric combinations were prepared using the hand lay-up method, considering the general design–construction characteristics. Further, an ultrasonic velocity decision method for ultrasonic inspection was proposed considering the GFRP hull design-construction characteristics, such as Gc and number of single-ply layers of hull plate by multiple linear regression method. The results show that the proposed method can reduce the thickness measurement error from approximately 20%–30% to 1%–2%, compared to an existing ultrasonic inspection method, only considering the Gc effect on ultrasonic velocity, which indicates that the proposed test method is suitable for practical applications.

WiKA: A Vision Based Sign Language Recognition from Extracted Hand Joint Features using DeepLabCut

The use of hand gesture recognition for sign language translation to address the communication gap between the hearing majority and the deaf population has had significant breakthroughs over the years. While the contact-based approach uses wearable devices, a vision-based solution is preferred owing to the convenience it offers and since it obviates the need for complicated gears. This study presents the development of WiKA, an open-source software designed to track the joints of the hands and interpret them into their corresponding sign language counterparts. DeepLabCut, a markerless pose estimation software, was employed to develop the Hand-Joint Tracking Model through the training of a sequential Convolutional Neural Network, utilizing extracted Hand-Joint features to predict the sign language alphabets (A-Z) and numbers (1-9) based on the positioning of the joints. The developed Hand-Joint tracking model exhibited a 4.92% training error and a 5.74% test error with a p-cut-off of 60%. On the other hand, the developed sign language recognition achieved a 96.44% prediction accuracy with only 0.0356% misclass. This model can be further integrated into mobile phones for seamless conversations between the signing and non-signing populations.

State-of-Health prediction of lithium-ion batteries based on a low dimensional Gaussian Process Regression

An accurate determination of the condition of a battery is a key challenge in operation. As the performance of lithium-ion batteries is degrading over time, an accurate prediction of the State-of-Health would improve the overall efficiency and safety. This paper presents a prediction method for the State-of-Health based on a Gaussian Process Regression with an automatic relevance determination kernel in a single model for three different types of battery cells. After reducing the dimension of the problem and a sensitivity analysis of the features, the model is trained, validated, and further tested on unseen data. A minimum test error is obtained with a mean absolute error of 1.33%. Combined with the low uncertainty of the prediction results, this shows the applicability and the great potential of forecasting the condition of a battery using data-driven methods.

GA-FCFNN: A new forecasting method combining feature selection methods and feedforward neural networks using genetic algorithms

In the modern landscape, the fusion of forecasting and computational intelligence empowers organizations to extract invaluable insights from vast datasets, facilitating informed decision-making, swift adaptation to market dynamics, and the enhancement of competitiveness, ultimately fostering innovation. Notably, forecasting has recently garnered significant attention, particularly within the realm of big data. This modern terrain presents a dual challenge: effectively distilling essential insights from complex data and optimizing data utilization. To address this challenge, we introduce a pioneering forecasting model (FM) engineered to excel in both information extraction and data utilization. Our approach commences with a formula designed to compute feature similarity, leveraging trend change data. These similarity metrics, applied to features and their relationships with labels, inform our feature selection process. This method integrates spatial and temporal correlations among features, fostering robust interconnections within the selected subset. Subsequently, we employ a feedforward neural network (FNN) to optimize feature subsets based on test error. While the FM trained with the optimal feature subset exhibits enhanced performance, it still shows some deviation from actual values in initial forecasting results (FORs). To mitigate these discrepancies, we implement a data utilization strategy. We refine the initial FORs by considering the contribution of each feature subset to them, thereby reducing disparities between forecasts and actual values. Model parameters are fine-tuned using a genetic algorithm (GA). Finally, we conduct an experimental analysis, comparing our FM to seven similar models using publicly available datasets. Our experimental results consistently demonstrate that our FM achieves high forecasting performance (FOP).

Next-generation sequencing survey of acute febrile illness in Senegal (2020-2022).

Acute febrile illnesses (AFI) in developing tropical and sub-tropical nations are challenging to diagnose due to the numerous causes and non-specific symptoms. The proliferation of rapid diagnostic testing and successful control campaigns against malaria have revealed that non-Plasmodium pathogens still contribute significantly to AFI burden. Thus, a more complete understanding of local trends and potential causes is important for selecting the correct treatment course, which in turn will reduce morbidity and mortality. Next-generation sequencing (NGS) in a laboratory setting can be used to identify known and novel pathogens in individuals with AFI. In this study, plasma was collected from 228 febrile patients tested negative for malaria at clinics across Senegal from 2020-2022. Total nucleic acids were extracted and converted to metagenomic NGS libraries. To identify viral pathogens, especially those present at low concentration, an aliquot of each library was processed with a viral enrichment panel and sequenced. Corresponding metagenomic libraries were also sequenced to identify non-viral pathogens. Sequencing reads for pathogens with a possible link to febrile illness were identified in 51/228 specimens, including (but not limited to): Borrelia crocidurae (N = 7), West Nile virus (N = 3), Rickettsia felis (N = 2), Bartonella quintana (N = 1), human herpesvirus 8 (N = 1), and Saffold virus (N = 1). Reads corresponding to Plasmodium falciparum were detected in 19 specimens, though their presence in the cohort was likely due to user error of rapid diagnostic testing or incorrect specimen segregation at the clinics. Mosquito-borne pathogens were typically detected just after the conclusion of the rainy season, while tick-borne pathogens were mostly detected before the rainy season. The three West Nile virus strains were phylogenetically characterized and shown to be related to both European and North American clades. Surveys such as this will increase the understanding of the potential causes of non-malarial AFI, which may help inform diagnostic and treatment options for clinicians who provide care to patients in Senegal.

Non-neurological factors associated with serum neurofilament levels in the United States population.

To model interdependencies of serum neurofilament light chain (sNfL), a clinically useful biomarker of axonal injury in neurological diseases, with demographic, anthropometric, physiological, and disease biomarkers in the United States population. sNfL and 80 biomarkers were obtained from the National Health and Nutrition Examination Survey (n = 2071, age: 20-75 years). Body habitus and composition, electrolytes, blood cell, metabolic, liver, and kidney function biomarkers, and common diseases were assessed with weighted regression adjusted for age, sex, and race/ethnicity. Salient biomarkers were modeled with ensemble learning; a Bayesian network structure was obtained for interdependencies. Age was strongly associated with sNfL. sNfL levels were 13% higher in men versus women. Mexican Americans had 18.5% lower sNfL versus Non-Hispanic Whites. sNfL was similar in pregnant versus nonpregnant women. Lymphocyte, and neutrophil numbers, and phosphorus, and chloride levels were associated with sNfL. Multiple liver function (e.g., albumin and gamma-glutamyltransferase), renal function (e.g., creatinine and urea), and carbohydrate/lipid metabolism markers (e.g., glucose and triglycerides) were associated with sNfL. A 50% greater creatinine was associated with 26.8% greater sNfL. Diabetes, kidney disease, congestive heart failure, and stroke were associated with sNfL. The ensemble learning algorithm predicted high sNfL outliers with 5.06%-9.16% test error. Bayesian network modeling indicated sNfL had neighbor dependencies with age, creatinine, albumin, and chloride. sNfL is associated with age, kidney and liver function, diabetes, blood cell subsets, and electrolytes. sNfL may be a useful biomarker for biological age of the whole body and major organ systems including the brain.

A Self-Defense Copyright Protection Scheme for NFT Image Art Based on Information Embedding

Non-convertible tokens (NFTs) have become a fundamental part of the metaverse ecosystem due to its uniqueness and immutability. However, existing copyright protection schemes of NFT image art relied on the NFTs itself minted by third-party platforms. A minted NFT image art only tracks and verifies the entire transaction process, but the legitimacy of the source and ownership of its mapped digital image art cannot be determined. The original author or authorized publisher lack an active defense mechanism to prove ownership of the digital image art mapped by the unauthorized NFT. Therefore, we propose a self-defense copyright protection scheme for NFT image art based on information embedding in this paper, called SDCP-IE. The original author or authorized publisher can embed the copyright information into the published digital image art without damaging its visual effect in advance. Different from the existing information embedding works, the proposed SDCP-IE can generally enhance the invisibility of copyright information with different embedding capacity. Furthermore, considering the scenario of copyright information being discovered or even destroyed by unauthorized parties, the designed SDCP-IE can efficiently generate enhanced digital image art to improve the security performance of embedded image, thus resisting the detection of multiple known and unknown detection models simultaneously. The experimental results have also shown that the PSNR values of enhanced embedded image are all over 57db on three datasets BOSSBase, BOWS2 and ALASKA#2. Moreover, compared with existing information embedding works, the enhanced embedded images generated by SDCP-IE reaches the best transferability performance on the advanced CNN-based detection models. When the target detector is the pre-trained SRNet at 0.4bpp, the test error rate of SDCP-IE at 0.4bpp on the evaluated detection model YeNet reaches 53.38%, which is 4.92%, 28.62% and 7.05% higher than that of the UTGAN, SPS-ENH and Xie-Model, respectively.

Revisiting the Nexus between Renewable Energy Consumption and Economic Growth in Nigeria: The VECM Approach

The research investigates the effect of renewable energy consumption on economic growth as well as analyzing the dynamic relationship between both variables from 1990 to 2022. Annual secondary data covering the period between 1990 and 2022 were used in the study. Data on real gross domestic product, renewable energy consumption (% of total final energy consumption), gross capital formation (% of GDP), labor force, total and trade (% of GDP) were sourced from World Development Indicator. The study applies Johansen co-integration test and Vector Error Correction Model (VECM) to examine the effect of renewable energy consumption on economic growth and other explanatory variables as well as using pairwise granger causality test to examine the causal relationship between both variables. The empirical evidence revealed long-term relationship between renewable energy consumption and economic growth. Evidence shows that there is no causality either unidirectional or bi-directional between both variables. In short run, renewable energy consumption revealed a significant positive impact on economic growth. This shows that renewable energy may stimulate economic growth, particularly as it has a greater short-term impact on economic growth than capital formation. The key policy implication drawn from the results indicate the need for investment in renewable energy technologies and infrastructure, which can help to increase the availability and affordability of renewable energy, improve workforce development in Nigeria renewable energy sector.

The Communication Conundrum: A Pilot Cross-Sectional Descriptive Examination of Family Nurse Perspectives Surrounding Patient Information Exchange During Interfacility Patient Transfers in Montana.

Background: Interfacility patient transfers are fraught with issues such as missed or ineffective communication in Montana given wide geographic distance between facilities and variance in resources. Inaccurate, absent, or delayed patient details may negatively affect patient outcomes and further result in duplicative testing and medication errors. Objective: The objective of this study was to describe the process of patient information communication during interfacility transfers as perceived by nurses practicing in Montana. Methods: The study design was a pilot cross-sectional descriptive approach. An online Qualtrics survey included demographic questions, two exploratory communication competence instruments, and four open-ended questions regarding communicating interfacility transfer patient information. Results: A total of 33 nurses completed the study, with the majority practicing at a critical access hospital (n = 15, 47%). Communication competence mean scores increased with dyad conversations, and a lack of standardized handoff tools was noted as a challenge. Nurses identified the following as barriers in the interfacility transfer handoff: incivility, amount of paperwork, interoperability issues, incomplete or outdated information, time, and resources. Implications for Practice: There is wide variability in current communication practices, ranging from verbal to electronic document transfers. The rural healthcare space is prime to continue examinations surrounding workflow optimization, accuracy, and consistency in shared information exchange at the time of interfacility transfer. There is an opportunity for potential training and education surrounding effective communication, interpersonal behaviors that support cross-organizational interactions, and the development of a standardized handoff tool contextual for interfacility transfer patients.

Exploring Cognitive Dysfunction in Long COVID Patients: Eye Movement Abnormalities and Frontal-Subcortical Circuits Implications via Eye-Tracking and Machine Learning

BackgroundCognitive dysfunction is regarded as one of the most severe aftereffects following coronavirus disease 2019 (COVID-19). Eye movements, controlled by several brain areas, such as the dorsolateral prefrontal cortex and frontal-thalamic circuits, provide a potential metric for assessing cortical networks and cognitive status. We aimed to examine the utility of eye movement measurements in identifying cognitive impairments in long COVID patients. MethodsWe recruited 40 long COVID patients experiencing subjective cognitive complaints and 40 healthy controls and used a certified eye-tracking medical device to record saccades and antisaccades. Machine learning was applied to enhance the analysis of eye movement data. ResultsPatients did not differ from the healthy controls regarding age, sex, and years of education. However, the patients' Montreal Cognitive Assessment total score was significantly lower than healthy controls. Most eye movement parameters were significantly worse in patients. These included the latencies, gain (computed as the ratio between stimulus amplitude and gaze amplitude), velocities, and accuracy (evaluated by the presence of hypermetric or hypometria dysmetria) of both visually and memory-guided saccades; the number of correct memory saccades; the latencies and duration of reflexive saccades; and the number of errors in the antisaccade test. Machine learning permitted distinguishing between long COVID patients experiencing subjective cognitive complaints and healthy controls. ConclusionOur findings suggest impairments in frontal subcortical circuits among long COVID patients who report subjective cognitive complaints. Eye-tracking, combined with machine learning, offers a novel, efficient way to assess and monitor long COVID patients' cognitive dysfunctions, suggesting its utility in clinical settings for early detection and personalized treatment strategies. Further research is needed to determine the long-term implications of these findings and the reversibility of cognitive dysfunctions.

Estimation of Proton Stopping Power Ratio and Mean Excitation Energy Using Electron Density and Its Applications via Machine Learning Approach.

The purpose of this study was to develop a simple flexible method for accurate estimation of stopping power ratio (SPR) and mean excitation energy (I) using relative electron density (ρ e). The model was formulated using empirical relationships between SPR, mean excitation energy I, and relative electron density. Some examples were implemented, and a comparison was carried out using other existing methods. The needed coefficients in the model were estimated using optimization tools. Basis vector method (BVM) and Hunemohr and Saito (H-S) method were applied to estimate the ρ e used in the application section. 80 kVp and 150 kVpSn were used as low and high energy, respectively, for the implementation of dual-energy methods. All the examples of the proposed method considered have modeling error that is ≤0.32% and testing root mean square error (RMSE) ≤0.92% for SPR with a mean error close to 0.00%. The method was able to achieve modeling RMSE of 2.12% for mean excitation energy with room for improvement. Similar or better results were achieved in application to BVM. The method showed robustness in application by achieving lower testing error than other presented methods in most cases. It achieved accurate estimation which can be improved using the machine learning algorithm since it is flexible to implement in terms of the function (model) degree and tissue classification.

Interest Rate Sensitivity of Depositors in Non-Financial Banks: A Case of an emerging Economy

This study uses 62 years of data, from 1960 to 2021, including variables like deposits, interest rates, GDP, and inflation, to investigate the link between interest rates and deposit behavior in non-Islamic banks in Pakistan. The study finds that interest rates have a big influence on deposit behavior using time series analytic techniques as unit root testing, cointegration analysis, and Vector Error Correction Model (VECM). More specifically, people's decisions to deposit money in non-Islamic banks are observed to be influenced by interest rate variations. The results highlight the importance of interest rate changes in influencing deposit behavior, showing how people choose which non-Islamic institutions to deposit money with. These findings have significance for financial institutions, policymakers, and economists in general. They provide important new information about the ways in which monetary policy choices affect deposit mobilization and liquidity conditions in the banking industry. The report also proposes directions for further investigation, including the possible examination of differences in deposit behavior between demographic groupings or geographical areas. To further help understand the dynamics at work, it also suggests examining how interest rates affect deposits in various sample configurations. This research advances theoretical frameworks in the field by shedding light on the complex interactions between interest rates and deposit behavior. It also offers stakeholders practical guidance on how to deal with the effects of interest rate dynamics on deposit mobilization and banking sector liquidity.

Preanalytical Errors in Clinical Laboratory Testing at a Glance: Source and Control Measures.

The accuracy of diagnostic results in clinical laboratory testing is paramount for informed healthcare decisions and effective patient care. While the focus has traditionally been on the analytical phase, attention has shifted towards optimizing the preanalytical phase due to its significant contribution to total laboratory errors. This review highlights preanalytical errors, their sources, and control measures to improve the quality of laboratory testing. Blood sample quality is a critical concern, with factors such as hemolysis, lipemia, and icterus leading to erroneous results. Sources of preanalytical errors encompass inappropriate test requests, patient preparation lapses, and errors during sample collection, handling, and transportation. Mitigating these errors includes harmonization efforts, education and training programs, automated methods for sample quality assessment, and quality monitoring. Collaboration between laboratory personnel and healthcare professionals is crucial for implementing and sustaining these measures to enhance the accuracy and reliability of diagnostic results, ultimately improving patient care.

Investigating the effectiveness of carbon nanotubes for the compressive strength of concrete using AI-aided tools

Sustainable development in the building industry can be achieved through the use of versatile cementitious composites. Thus, incorporating nanoparticles into cement composites can create materials with enhanced performance and numerous applications. The utilization of carbon nanotubes (CNTs) in the construction industry has great promise for developing efficient solutions to establish a sustainable ecosystem with diverse characteristics. However, forecasting the characteristics of these composites is a significant challenge due to their intricate composite structure and nonlinear behavior. Designing and conducting laboratory experiments on diverse samples and across multiple age groups is challenging, time-consuming, and costly. Moreover, there is presently a lack of a forecasting model that can predict concrete's compressive strength (fc') with nanoparticles. Three machine learning (ML) techniques, K-nearest neighbor (KNN), linear regression (LR), and artificial neural network (ANN), were used to predict the fc' of nanocomposites containing CNTs in this research. A thorough database consisting of 282 data entities for the CNTs-based concrete and the model's reliability was assessed using the R2 test and statistical error analysis. The ANN model had the most outstanding R2 value of 0.885, while the KNN and LR models had R2 values of 0.838 and 0.744, respectively. Moreover, RReliefF analysis is utilized to evaluate the primary components in predicting concrete outcomes. This research shows that the properties of CNT-based concrete composites are greatly affected by the water-to-binder ratio, followed by the proportions of cement and coarse aggregates. The ML algorithms exhibited superior generalization capabilities, suggesting their enhanced potential for accurate predictions of CNTs-based concrete properties.

The effects of neuro-vestibular-ocular exercises and myofascial release on proprioception and performance in football players with chronic ankle instability

BACKGROUND AND PURPOSE: Football players often use quick change of direction, sudden acceleration and explosive movements. This is why chronic ankle instability is one of the most common conditions affecting this athlete population. This study investigates the effects of neuro-vestibular-ocular exercises and myofascial release on proprioception and performance in football players with chronic ankle instability. MATERIALS AND METHODS: This randomized controlled clinical trial included 60 football players aged 18–30. The players were randomly divided into two groups. The first group (NVOEG, n=30) was included in the Neuro-Vestibular-ocular exercise training program, while the myofascial release was applied to the second group (MRHOG, n=30). Both protocols were applied for eight weeks. The participants’ Cumberland Ankle Instability Tool (CAIT) scores were assessed before and after the intervention. Proprioception was evaluated by using the joint position sense test. Additionally, kick speed (KS), 30-m sprint, zig-zag agility test (ZAT) and Landing error score system (LESS) tests were applied to evaluate their performance. RESULTS: Post-rehabilitation scores showed significant differences in proprioception and performance parameters compared to pre-scores in both groups. (p<0.05). Significant differences were observed between the NVOEG and MRHOG in the post-test scores, including CAIT for unstable and stable ankle, proprioception for unstable ankle, LESS, ZAT with ball performance, and 30-meter sprint test. CONCLUSIONS: The neuro-vulvular-ocular exercise training protocol had superior results in terms of proprioception and performance parameters in chronic ankle instability management for football players.

Behavioral carry-over effect and power consideration in crossover trials.

A crossover trial is an efficient trial design when there is no carry-over effect. To reduce the impact of the biological carry-over effect, a washout period is often designed. However, the carry-over effect remains an outstanding concern when a washout period is unethical or cannot sufficiently diminish the impact of the carry-over effect. The latter can occur in comparative effectiveness research, where the carry-over effect is often non-biological but behavioral. In this paper, we investigate the crossover design under a potential outcomes framework with and without the carry-over effect. We find that when the carry-over effect exists and satisfies a sign condition, the basic estimator underestimates the treatment effect, which does not inflate the type I error of one-sided tests but negatively impacts the power. This leads to a power trade-off between the crossover design and the parallel-group design, and we derive the condition under which the crossover design does not lead to type I error inflation and is still more powerful than the parallel-group design. We also develop covariate adjustment methods for crossover trials. We evaluate the performance of cross-over design and covariate adjustment using data from the MTN-034/REACH study.

Nitrogen dioxide exposure, attentional function, and working memory in children from 4 to 8 years: Periods of susceptibility from pregnancy to childhood

BackgroundAir pollution exposure during pregnancy and childhood has been linked to executive function impairment in children, however, very few studies have assessed these two exposure periods jointly to identify susceptible periods of exposure. We sought to identify potential periods of susceptibility of nitrogen dioxide (NO2) exposure from conception to childhood on attentional function and working memory in school-aged children. MethodsWithin the Spanish INMA Project, we estimated residential daily NO2 exposures during pregnancy and up to 6 years of childhood using land use regression models (n = 1,703). We assessed attentional function at 4–6 years and 6–8 years, using the Conners Kiddie Continuous Performance Test and the Attention Network Test, respectively, and working memory at 6–8 years, using the N-back task. We used distributed lag non-linear models to assess the periods of susceptibility of each outcome, adjusting for potential confounders and correcting for multiple testing. We also stratified all models by sex. ResultsHigher exposure to NO2 between 1.3 and 1.6 years of age was associated with higher hit reaction time standard error (HRT-SE) (0.14 ms (95 % CI 0.05; 0.22) per 10 μg/m3 increase in NO2) and between 1.5 and 2.2 years of age with more omission errors (1.02 (95 % CI 1.01; 1.03) of the attentional function test at 4–6 years. Higher exposure to NO2 between 0.3 and 2.2 years was associated with higher HRT-SE (10.61 ms (95 % CI 3.46; 17.75) at 6–8 years only in boys. We found no associations between exposure to NO2 and working memory at 6–8 years. ConclusionOur findings suggest that NO2 exposure during the first two years of life is associated with poorer attentional function in children from 4 to 8 years of age, especially in boys. These findings highlight the importance of exploring long-term effects of traffic-related air pollution exposure in older age groups.

Investigation of the Effects of Caffeine Intake on Reaction Time, Acceleration, Speed and Balance in Short Distance Runners

The aims of this study is to examine the effects of caffeine intake in sprinters on reaction time, speed, acceleration and balance. A total of 31 sprinters (18 female, 13 male) from the athletics team of university faculty of sports sciences participated in the study. Sprinters were divided into two groups, caffeine and placebo, and their balance, reaction time, 30-meter running time and 15-meter acceleration performances were measured. The Wilcoxon t-Test and the Mann-Whitney U test were applied to the data obtained on the SPSS 23.0 package program. As a result of the study, the number of errors in the balance test decreased significantly in both the experimental group and the control group while a significant decrease was observed in block takeoff reaction time only in the experimental group. In conclusion, it was determined that while caffeine intake did not have a significant effect on acceleration and running times, block take off reaction time created a significant difference in the experimental group and could provide an advantage to the sprinters. Therefore, it can be said that caffeinated gum may be a good ergogenic aid to improve reaction performance in sprinters.

Effects of feature reduction on emotion recognition using EEG signals and machine learning

AbstractElectroencephalography is a core technology of brain computer interfaces. Even a few number of electrodes can produce complex signals that are difficult to interpret. This is particularly true when trying to detect complex mental states, such as the identification of human emotions. This work analyzes the impact of feature reduction using the SJTU Emotion EEG data set, considering inter‐subject and inter‐session EEG classification. It is well established that EEG data tends to be subject dependent, and that the brain patterns produced by a person in response to a specific stimuli will tend to change over their lifetime, we can say that the above classification task is one of the most challenging test scenarios. We extract 1068 time and frequency domain features, single and multiple channels. Feature reduction was analyzed using both filter and wrapper based approaches. In the first group we use principal component analysis, Kernel PCA and Locally Linear Embedding, and in the second three meta‐heuristic algorithms are employed. For classification, we apply multi‐layer perceptron, quadratic discriminant analysis, random forest and support vector machines. Results show up to 100% accuracy on the median test error for some splits in the 10‐fold cross‐validation process. Best results were achieved by filter‐based selection, with the best average accuracy reached by Random Forest (93.20%) and Multi‐Layer Perceptron (92.71%), using components found with Kernel PCA. To the best of our knowledge, these classification results are the best ever achieved for the SEED data set on the subject‐independent and session‐independent problem. Comparisons using a Leave‐One‐Subject‐Out cross‐validation setting shows that the proposed methodology achieves state‐of‐the‐art results using only 12 electrodes. These off‐the‐shelf ML techniques, coupled with filter based feature reduction, achieve the best results, outperforming complex classification models and wrapper based meta‐heuristics.