Machine Learning Analytics Research Articles

Predicting futility of upfront surgery in perihilar cholangiocarcinoma: Machine learning analytics model to optimize treatment allocation.

While resection remains the only curative option for perihilar cholangiocarcinoma, it is well known that such surgery is associated with a high risk of morbidity and mortality. Nevertheless, beyond facing life-threatening complications, patients may also develop early disease recurrence, defining a "futile" outcome in perihilar cholangiocarcinoma surgery. The aim of this study is to predict the high-risk category (futile group) where surgical benefits are reversed and alternative treatments may be considered. The study cohort included prospectively maintained data from 27 Western tertiary referral centers: the population was divided into a development and a validation cohort. The Framingham Heart Study methodology was used to develop a preoperative scoring system predicting the "futile" outcome. A total of 2271 cases were analyzed: among them, 309 were classified within the "futile group" (13.6%). American Society of Anesthesiology (ASA) score ≥ 3 (OR 1.60; p = 0.005), bilirubin at diagnosis ≥50mmol/L (OR 1.50; p = 0.025), Ca 19-9 ≥ 100U/mL (OR 1.73; p = 0.013), preoperative cholangitis (OR 1.75; p = 0.002), portal vein involvement (OR 1.61; p = 0.020), tumor diameter ≥3cm (OR 1.76; p < 0.001), and left-sided resection (OR 2.00; p < 0.001) were identified as independent predictors of futility. The point system developed, defined three (ie, low, intermediate, and high) risk classes, which showed good accuracy (AUC 0.755) when tested on the validation cohort. The possibility to accurately estimate, through a point system, the risk of severe postoperative morbidity and early recurrence, could be helpful in defining the best management strategy (surgery vs. nonsurgical treatments) according to preoperative features.

Survey and Evaluation of Hypertension Machine Learning Research.

Background Machine learning (ML) is pervasive in all fields of research, from automating tasks to complex decision-making. However, applications in different specialities are variable and generally limited. Like other conditions, the number of studies employing ML in hypertension research is growing rapidly. In this study, we aimed to survey hypertension research using ML, evaluate the reporting quality, and identify barriers to ML's potential to transform hypertension care. Methods and Results The Harmonious Understanding of Machine Learning Analytics Network survey questionnaire was applied to 63 hypertension-related ML research articles published between January 2019 and September 2021. The most common research topics were blood pressure prediction (38%), hypertension (22%), cardiovascular outcomes (6%), blood pressure variability (5%), treatment response (5%), and real-time blood pressure estimation (5%). The reporting quality of the articles was variable. Only 46% of articles described the study population or derivation cohort. Most articles (81%) reported at least 1 performance measure, but only 40% presented any measures of calibration. Compliance with ethics, patient privacy, and data security regulations were mentioned in 30 (48%) of the articles. Only 14% used geographically or temporally distinct validation data sets. Algorithmic bias was not addressed in any of the articles, with only 6 of them acknowledging risk of bias. Conclusions Recent ML research on hypertension is limited to exploratory research and has significant shortcomings in reporting quality, model validation, and algorithmic bias. Our analysis identifies areas for improvement that will help pave the way for the realization of the potential of ML in hypertension and facilitate its adoption.

Post-Covid Remote Patient Monitoring using Medical Internet of Things and Machine Learning Analytics

The Covid-19 pandemic disturbed the smooth functioning of healthcare services throughout the world. New practices such as masking, social distancing and so on were followed to prevent the spread. Further, the severity of the problem increases for the elderly people and people having co-morbidities as proper medical care was not possible and as a result many deaths were recorded. Even for those patients who recovered from Covid could not get proper health monitoring in the Post-Covid phase as a result many deaths and severity in health conditions were reported after the Covid recovery i.e., the Post-Covid era. Technical interventions like the Internet of Things (IoT) based remote patient monitoring using Medical Internet of Things (M-IoT) wearables is one of the solutions that could help in the Post-Covid scenarios. The paper discusses a proposed framework where in a variety of IoT sensing devices along with ML algorithms are used for patient monitoring by utilizing aggregated data acquired from the registered Post-Covid patients. Thus, by using M-IoT along with Machine Learning (ML) approaches could help us in monitoring Post-Covid patients with co-morbidities for and immediate medical help.

A clinical decision support system for AI-assisted decision-making in response-adaptive radiotherapy (ARCliDS)

Involvement of many variables, uncertainty in treatment response, and inter-patient heterogeneity challenge objective decision-making in dynamic treatment regime (DTR) in oncology. Advanced machine learning analytics in conjunction with information-rich dense multi-omics data have the ability to overcome such challenges. We have developed a comprehensive artificial intelligence (AI)-based optimal decision-making framework for assisting oncologists in DTR. In this work, we demonstrate the proposed framework to Knowledge Based Response-Adaptive Radiotherapy (KBR-ART) applications by developing an interactive software tool entitled Adaptive Radiotherapy Clinical Decision Support (ARCliDS). ARCliDS is composed of two main components: Artifcial RT Environment (ARTE) and Optimal Decision Maker (ODM). ARTE is designed as a Markov decision process and modeled via supervised learning. Given a patient’s pre- and during-treatment information, ARTE can estimate treatment outcomes for a selected daily dosage value (radiation fraction size). ODM is formulated using reinforcement learning and is trained on ARTE. ODM can recommend optimal daily dosage adjustments to maximize the tumor local control probability and minimize the side effects. Graph Neural Networks (GNN) are applied to exploit the inter-feature relationships for improved modeling performance and a novel double GNN architecture is designed to avoid nonphysical treatment response. Datasets of size 117 and 292 were available from two clinical trials on adaptive RT in non-small cell lung cancer (NSCLC) patients and adaptive stereotactic body RT (SBRT) in hepatocellular carcinoma (HCC) patients, respectively. For training and validation, dense data with 297 features were available for 67 NSCLC patients and 110 features for 71 HCC patients. To increase the sample size for ODM training, we applied Generative Adversarial Networks to generate 10,000 synthetic patients. The ODM was trained on the synthetic patients and validated on the original dataset. We found that, Double GNN architecture was able to correct the nonphysical dose-response trend and improve ARCliDS recommendation. The average root mean squared difference (RMSD) between ARCliDS recommendation and reported clinical decisions using double GNNs were 0.61 [0.03] Gy/frac (mean [sem]) for adaptive RT in NSCLC patients and 2.96 [0.42] Gy/frac for adaptive SBRT HCC compared to the single GNN’s RMSDs of 0.97 [0.12] Gy/frac and 4.75 [0.16] Gy/frac, respectively. Overall, For NSCLC and HCC, ARCliDS with double GNNs was able to reproduce 36% and 50% of the good clinical decisions (local control and no side effects) and improve 74% and 30% of the bad clinical decisions, respectively. In conclusion, ARCliDS is the first web-based software dedicated to assist KBR-ART with multi-omics data. ARCliDS can learn from the reported clinical decisions and facilitate AI-assisted clinical decision-making for improving the outcomes in DTR.

Real-time insight into the multistage mechanism of nanoparticle exsolution from a perovskite host surface

In exsolution, nanoparticles form by emerging from oxide hosts by application of redox driving forces, leading to transformative advances in stability, activity, and efficiency over deposition techniques, and resulting in a wide range of new opportunities for catalytic, energy and net-zero-related technologies. However, the mechanism of exsolved nanoparticle nucleation and perovskite structural evolution, has, to date, remained unclear. Herein, we shed light on this elusive process by following in real time Ir nanoparticle emergence from a SrTiO3 host oxide lattice, using in situ high-resolution electron microscopy in combination with computational simulations and machine learning analytics. We show that nucleation occurs via atom clustering, in tandem with host evolution, revealing the participation of surface defects and host lattice restructuring in trapping Ir atoms to initiate nanoparticle formation and growth. These insights provide a theoretical platform and practical recommendations to further the development of highly functional and broadly applicable exsolvable materials.

The presence, nature and network characteristics of behavioural phenotypes in temporal lobe epilepsy.

The relationship between temporal lobe epilepsy and psychopathology has had a long and contentious history with diverse views regarding the presence, nature and severity of emotional-behavioural problems in this patient population. To address these controversies, we take a new person-centred approach through the application of unsupervised machine learning techniques to identify underlying latent groups or behavioural phenotypes. Addressed are the distinct psychopathological profiles, their linked frequency, patterns and severity and the disruptions in morphological and network properties that underlie the identified latent groups. A total of 114 patients and 83 controls from the Epilepsy Connectome Project were administered the Achenbach System of Empirically Based Assessment inventory from which six Diagnostic and Statistical Manual of Mental Disorders-oriented scales were analysed by unsupervised machine learning analytics to identify latent patient groups. Identified clusters were contrasted to controls as well as to each other in order to characterize their association with sociodemographic, clinical epilepsy and morphological and functional imaging network features. The concurrent validity of the behavioural phenotypes was examined through other measures of behaviour and quality of life. Patients overall exhibited significantly higher (abnormal) scores compared with controls. However, cluster analysis identified three latent groups: (i) unaffected, with no scale elevations compared with controls (Cluster 1, 37%); (ii) mild symptomatology characterized by significant elevations across several Diagnostic and Statistical Manual of Mental Disorders-oriented scales compared with controls (Cluster 2, 42%); and (iii) severe symptomatology with significant elevations across all scales compared with controls and the other temporal lobe epilepsy behaviour phenotype groups (Cluster 3, 21%). Concurrent validity of the behavioural phenotype grouping was demonstrated through identical stepwise links to abnormalities on independent measures including the National Institutes of Health Toolbox Emotion Battery and quality of life metrics. There were significant associations between cluster membership and sociodemographic (handedness and education), cognition (processing speed), clinical epilepsy (presence and lifetime number of tonic-clonic seizures) and neuroimaging characteristics (cortical volume and thickness and global graph theory metrics of morphology and resting-state functional MRI). Increasingly dispersed volumetric abnormalities and widespread disruptions in underlying network properties were associated with the most abnormal behavioural phenotype. Psychopathology in these patients is characterized by a series of discrete latent groups that harbour accompanying sociodemographic, clinical and neuroimaging correlates. The underlying neurobiological patterns suggest that the degree of psychopathology is linked to increasingly dispersed abnormal brain networks. Similar to cognition, machine learning approaches support a novel developing taxonomy of the comorbidities of epilepsy.

Identification of PsA phenotypes with machine learning analytics using data from two phase III clinical trials of guselkumab in a bio-naïve population of patients with PsA

ObjectivesPsoriatic arthritis (PsA) phenotypes are typically defined by their clinical components, which may not reflect patients’ overlapping symptoms. This post hoc analysis aimed to identify hypothesis-free PsA phenotype clusters using...

Characterisation of Composite Materials for Wind Turbines Using Frequency Modulated Continuous Wave Sensing

Wind turbine blades (WTBs) are critical sub-systems consisting of composite multi-layer material structures. WTB inspection is a complex and labour intensive process, and failure of it can lead to substantial energy and economic losses to asset owners. In this paper, we proposed a novel non-destructive evaluation method for blade composite materials, which employs Frequency Modulated Continuous Wave (FMCW) radar, robotics and machine learning (ML) analytics. We show that using FMCW raster scan data, our ML algorithms (SVM, BP, Decision Tree and Naïve Bayes) can distinguish different types of composite materials with accuracy of over 97.5%. The best performance is achieved by SVM algorithms, with 94.3% accuracy. Furthermore, the proposed method can also achieve solid results for detecting surface defect: interlaminar porosity with 80% accuracy overall. In particular, the SVM classifier shows highest accuracy of 92.5% to 98.9%. We also show the ability to detect air voids of 1mm differences within the composite material WT structure with 94.1% accuracy performance using SVM, and 84.5% using Naïve Bayes. Lastly, we create a digital twin of the physical composite sample to support the integration and qualitative analysis of the FMCW data with respect to composite sample characteristics. The proposed method explores a new sensing modality for non-contact surface and subsurface for composite materials, and offer insights for developing alternative, more cost-effective inspection and maintenance regimes.

State-of-the-art Applications of Patient-Reported Outcome Measures in Spinal Care.

Patient-reported outcome measures (PROMs) assign objective measures to patient's subjective experiences of health, pain, disability, function, and quality of life. PROMs can be useful for providers in shared decision making, outcome assessment, and indicating patients for surgery. In this article, we provide an overview of the legacy PROMs used in spinal care, recent advancements in patient-reported outcomes, and future directions in PROMs. Recent advances in patient-reported outcome assessments have included standardization of measurement tools, integration of data collection into workflow, and applications of outcome measures in predictive models and decision-making tools. Continual appraisal of instruments and incorporation into artificial intelligence and machine learning analytics will continue to augment the delivery of high-value spinal care.

Prediction of Ovarian Cancer with Deep Machine Learning and Alternative Splicing

Objective: Early detection of ovarian cancer could lead to improved survival rates, however no method has reliably been able to predict ovarian cancer. The aim of this study is to determine if processing alternative splicing data from high grade serous ovarian cancer patients using machine learning analytics will discriminate high grade serous ovarian cancer from normal fallopian tube samples. The ultimate goal would be to have a model that can predict high grade serous ovarian cancer with a blood test. Methods: This is a case-control study of patients with confirmed high grade serous ovarian cancer and those undergoing salpingectomy for benign indications. RNA-sequencing was performed on all samples. RNA-sequence data was then put into Deep-learning augmented RNA-seq analysis of transcript splicing software suite. Deep-learning augmented RNA-seq analysis of transcript splicing created a model of differential alternative splicing aimed to discriminate between high grade serous ovarian cancer and normal fallopian tube. DEXSeq analysis was used to determine exon-based expression. Initial results with both analytics were then modelled with multivariate lasso regression to create prediction models (performance determined by area under the curve and 95% CI). Models created were the validated using The Cancer Genome Atlas data sets. Results: One hundred and twelve high grade serous ovarian cancer and 12 benign samples were successfully sequenced. Deep-learning augmented RNA-sequencing analysis of transcript splicing identified 998 unique differentially expressed exons between high grade serous ovarian cancer and controls. Multivariate lasso regression analysis identified several exons that predicted high grade serous ovarian cancer with high performance. Specifically, ENSG00000182512:E001 from gene GLRX5 was highly predictive of high grade serous ovarian cancer with an area under the curve of 100%. Conclusions: Application of machine learning analytics to exon differential expression, most likely due to alternative splicing, predicted high grade serous ovarian cancer with high performance. These results were validated in an independent dataset of cases and controls. Differential exon expression from cell-free RNA potentially could be used for early diagnosis of high grade serous ovarian cancer.

Rapid Detection of Synthetic Cannabinoid Receptor Agonists Impregnated into Paper by Raman Spectroscopy

Abstract English The last decade has witnessed the emergence of new psychoactive substances that are analogues of classical drugs of abuse in order to escape the regulations surrounding the latter drugs. These drugs were of both herbal and synthetic origin and were advertised initially as ‘legal highs’; thus, they were perceived as safe by users. Hence, upon their emergence, they were not controlled by the Misuse of Drugs Act 1971, which contributed to their popularity and increased sales online and within street markets. In 2016, the Psychoactive Substance Act introduced a blanket ban on all new psychoactive substances except for caffeine, alcohol, and nicotine. This in turn, contributed to the change in the sale of new psychoactive substances products that have been sold as concealed in different matrices, including herbal products, papers, fabrics, and textiles. Concealing drugs in paper has been very popular, especially since the drug product is of lightweight and can be sent via postal services. However, new psychoactive concealed in papers are toxic not only to the users; but also, to the person handling them (i.e. mail employees). One of the classes of new psychoactive substances that have been commonly concealed in papers and that have been linked to toxicity and hospitalization cases is synthetic cannabinoids. Therefore, there is a need to identify new psychoactive substances concealed in papers non-destructively and rapidly to prevent toxicity linked to them. Handheld Raman spectroscopy offers this advantage as it is of lightweight and carries the sample to the matrix. Therefore, this work used handheld Raman spectroscopy for identifying synthetic cannabinoids concealed in papers using Raman spectroscopy combined with machine learning analytics. Synthetic cannabinoid and paper samples were measured non-destructively using a handheld Raman spectrometer equipped with a 1064 nm laser wavelength. Spectral data was exported into Matlab 2020b where machine learning analytics including identification and prediction was. The results showed that Raman spectroscopy could identify specific synthetic cannabinoids in papers that were either deposited on the surface of the paper or diffused inside the paper substrate. When machine learning analytics were applied to the Raman spectra of the papers, quantitative information was obtained regarding the amount of synthetic cannabinoid deposited on the paper surface. In summary, handheld Raman spectroscopy could identify and quantify synthetic cannabinoids on paper rapidly and non-destructively. Future work involves testing other classes of new psychoactive substance and applying deep learning analytics

Accuracy of Supervised Machine Learning in Predicting Depression, Anxiety and Stress Using Web-based Big Data: Preserving the Humanistic Intellect

Introduction: One of the most useful tool to assess the extent of depression, anxiety and stress symptoms is the validated Depression, Anxiety and Stress Scale, 21 items (DASS-21). The availability of online mental health resource centre provides big data capable of machine learning analytics for early detection of mental health issues. However, prediction accuracy of these data using machine learning method remains elusive. Methods: A cross sectional study was conducted, using secondary data of respondents who answered an online DASS-21 questionnaire from an online resource center. Depression, anxiety and stress were measured using DASS21 as either the outcome or predictor, depending on the model. The model includes sociodemographic predictors such as gender, age, race, marital status, education level and occupational status. A feed-forward artificial neural network was constructed based on multilayer perceptron machine learning procedure using IBM SPSS version 23. Results: A total of 339,781 respondents data were obtained. The observed prevalence of depression, anxiety and stress was 39.9%, 48.5% and 13.4%, respectively. This resulted in 76.4% prediction accuracy for depression, 76.3% accuracy for anxiety and 87.4% prediction accuracy for stress. Stress and anxiety were the most important factors contributing to the disease model. Conclusion: The prediction models have high accuracy to predict the true observed depression, anxiety and stress prevalence. The clinical relevance of these prediction models still needs the human intellect judgment based on Maqasid al-Shariah principles. Machine learning therefore should not be abused but to help in decision-making towards early detection and prompt treatment.

Machine Learning Analytics for Predicting Tax Revenue Potential

In line with rapid business process digitalization in the Directorate General of Taxes, the size of the data stored in the institution has grown exponentially. However, there is a problem with generating value out of the valuable data assets. Correspondingly, this research provides machine-learning-based predictive analytics as a solution to the question of how to use taxpayers' trigger data as a decision support system to discover and realize unexplored tax potential. More specifically, this research presents predictive analytics models that can accurately predict which potential taxpayers are likely to pay their due. We developed three machine learning models: logistic regression, random forest, and decision tree. We analyzed 5,562 tax revenue potential data samples with eight predictors: trigger data nominal value, distance to tax office, type of taxpayer, media of tax report, type of tax, report status, registered year of taxpayer, and area coverage. Our study shows that the random forest model provided the best prediction performance. The resultant weight of each attribute indicated that the status of the tax report was the top tier of variable importance in predicting tax revenue potential. The analytics can help tax officers determine potential taxpayers with the highest likelihood to pay their due. Given the size of the data records, this approach can provide tax administrators with a powerful tool to increase work efficiency, combat tax evasion, and provide better customer service.

Literature review of the reciprocal value of artificial and human intelligence in early childhood education

This paper presents a systematic literature review of artificial intelligence (AI)-supported teaching and learning in early childhood. The focus is on human–machine cooperation in education. International evidence and associated problems with the reciprocal contributions of humans and machines are presented and discussed, as well as future horizons regarding AI research in early education. Also, the ethical implications of applying machine learning, deep learning and learning analytics in early childhood education are considered. The method adopted has five steps: identification of the research, evaluation and selection of the literature, data extraction, synthesis, and results. The results shown that AI applications still present limitations in terms of the challenges encountered in early childhood education and data privacy and protection policies.

Validation of a Mobile, Sensor-based Neurobehavioral Assessment With Digital Signal Processing and Machine-learning Analytics.

The Miro Health Mobile Assessment Platform consists of self-administered neurobehavioral and cognitive assessments that measure behaviors typically measured by specialized clinicians. To evaluate the Miro Health Mobile Assessment Platform's concurrent validity, test-retest reliability, and mild cognitive impairment (MCI) classification performance. Sixty study participants were evaluated with Miro Health version V.2. Healthy controls (HC), amnestic MCI (aMCI), and nonamnestic MCI (naMCI) ages 64-85 were evaluated with version V.3. Additional participants were recruited at Johns Hopkins Hospital to represent clinic patients, with wider ranges of age and diagnosis. In all, 90 HC, 21 aMCI, 17 naMCI, and 15 other cases were evaluated with V.3. Concurrent validity of the Miro Health variables and legacy neuropsychological test scores was assessed with Spearman correlations. Reliability was quantified with the scores' intraclass correlations. A machine-learning algorithm combined Miro Health variable scores into a Risk score to differentiate HC from MCI or MCI subtypes. In HC, correlations of Miro Health variables with legacy test scores ranged 0.27-0.68. Test-retest reliabilities ranged 0.25-0.79, with minimal learning effects. The Risk score differentiated individuals with aMCI from HC with an area under the receiver operator curve (AUROC) of 0.97; naMCI from HC with an AUROC of 0.80; combined MCI from HC with an AUROC of 0.89; and aMCI from naMCI with an AUROC of 0.83. The Miro Health Mobile Assessment Platform provides valid and reliable assessment of neurobehavioral and cognitive status, effectively distinguishes between HC and MCI, and differentiates aMCI from naMCI.

Recent advances in the use of machine learning and artificial intelligence to improve diagnosis, predict flares, and enrich clinical trials in lupus.

Machine learning is a computational tool that is increasingly used for the analysis of medical data and has provided the promise of more personalized care. The frequency with which machine learning analytics are reported in lupus research is comparable with that of rheumatoid arthritis and cancer, yet the clinical application of these computational tools has yet to be translated into better care. Considerable work has been applied to the development of machine learning models for lupus diagnosis, flare prediction, and classification of disease using histology or other medical images, yet few models have been tested in external datasets and independent centers. Application of machine learning has yet to be reported for lupus clinical trial enrichment and automated identification of eligible patients. Integration of machine learning into lupus clinical care and clinical trials would benefit from collaborative development between clinicians and data scientists. Although the application of machine learning to lupus data is at a nascent stage, initial results suggest a promising future.

Actinides in complex reactive media: A combined ab initio molecular dynamics and machine learning analytics study of transuranic ions in molten salts

The predominant ionic chemistry and the similarity in ionic radius of actinides make it very difficult to structurally distinguish them in liquids. To tackle this problem, we investigate actinides in molten salts using ab initio molecular dynamics and machine learning analytics. Ab initio simulations show that the f-states clearly affects the electronic properties while their impact on structural properties is not obvious. For the series of trivalent actinides U3+, Pu3+, Cm3+, Cf3+, and Fm3+ in molten NaCl and FLiBe, actinide-ligand bonds have a higher degree of covalency in NaCl (than in FLiBe), and a higher degree of ionicity in FLiBe. Furthermore, a machine learned classification model can distinguish atomic environments of chemically similar actinides with more than 80% confidence, as long as atoms beyond the first solvation shells are considered. Our work shows that only two types of descriptors are necessary to account for all the fluctuations in heavy metal/molten salt mixtures: The first descriptor represents the electronic state of the heavy metal, while the second encompasses the local coordination environment.

Gender differences in serious police misconduct: A machine-learning analysis of the New York Police Department (NYPD)

PurposeDespite a considerable body of research on police misconduct, findings have been mixed, with little consensus regarding its causes and best practices for prevention. Emerging research has focused on the role of gender in understanding and preventing misconduct. The current study examines the extent to which the features associated with serious misconduct differ between male and female officers. MethodsUsing a unique complaint dataset from the NYPD, we apply a sequence of machine learning analytics to consider if it is possible to predict serious misconduct among either group, and whether key predictors differ between groups. ResultsThe results show that it was possible to predict serious misconduct among each group with considerable confidence, while there were notable differences in prevalence, and type of misconduct between sexes. ConclusionsFindings hold important implications for policy, prevention and analytical approaches to police misconduct.

Machine learning analytics for virtual bidding in the electricity market

In order to solve the problem of the high risks and low efficiency caused by the inconsistency of the day-ahead and real-time prices in two-settlement electricity market, virtual bidding is used to arbitrage on the difference between such two market prices that are unknown to virtual bidders to promote the price convergence. The problem of optimal bidding for virtual bidders from the spatio-temporal dimensions is addressed in this paper. The model takes the budget constraints of virtual bidders into account, as well as considers decrement and increment bids of virtual bidding to maximize the cumulative payoff of virtual bidders, which is formulated as a Markov Decision Process problem. Meanwhile, the conditional value-at-risk is used to quantify and hedge the risks faced by virtual bidders. A deep reinforcement learning algorithm is used to achieve an effective solution to the optimal bidding strategy problem through continuous interaction with a simulated building environment to obtain feedback and update the parameters of the neural network without referring to any prior model knowledge. The PJM data from 2016 to 2018 is used to calculate the cumulative profits and Sharpe ratio of virtual bidders. Compared with greedy algorithm and dynamic programming, the deep reinforcement learning algorithm is verified the effectiveness and superiority in this paper.

Development and Integration of Metocean Data Interoperability for Intelligent Operations and Automation Using Machine Learning: A Review

The current oil industry is moving towards digitalization, which is a good opportunity that will bring value to all its stakeholders. The digitalization of oil and gas discovery, which are production-based industries, is driven by enabling technologies which include machine learning (ML) and big data analytics. However, the existing Metocean system generates data manually using sensors such as the wave buoy, anemometer, and acoustic doppler current profiler (ADCP). Additionally, these data which appear in ASCII format to the Metocean system are also manual and silos. This slows down provisioning, while the monitoring element of the Metocean data path is partial. In this paper, we demonstrate the capabilities of ML for the development of Metocean data integration interoperability based on intelligent operations and automation. A comprehensive review of several research studies, which explore the needs of ML in oil and gas industries by investigating the in-depth integration of Metocean data interoperability for intelligent operations and automation using an ML-based approach, is presented. A new model integrated with the existing Metocean data system using ML algorithms to monitor and interoperate with maximum performance is proposed. The study reveals that ML is one of the crucial and key enabling tools that the oil and gas industries are now focused on for implementing digital transformation, which allows the industry to automate, enhance production, and have less human capacity. Lastly, user recommendations for potential future investigations are offered.