Search

Causal machine learning (ML) algorithms recover graphical structures that tell us something about cause-and-effect relationships. The causal representation provided by these algorithms enables transparency and explainability, which is necessary for decision making in critical real-world problems. Yet, causal ML has had limited impact in practice compared to associational ML. This paper investigates the challenges of causal ML with application to COVID-19 UK pandemic data. We collate data from various public sources and investigate what the various structure learning algorithms learn from these data. We explore the impact of different data formats on algorithms spanning different classes of learning, and assess the results produced by each algorithm, and groups of algorithms, in terms of graphical structure, model dimensionality, sensitivity analysis, confounding variables, predictive and interventional inference. We use these results to highlight open problems in causal structure learning and directions for future research. To facilitate future work, we make all graphs, models, data sets, and source code publicly available online. [ABSTRACT FROM AUTHOR]

• We studied 72,363 cyclist crashes that occurred in Great Britain in the period 2016–2018. • We investigated factors contributing to higher proportion of severe and fatal crashes. • We used both a machine learning method as well as an econometric model. • We found unobserved heterogeneity in the data. • The models resulted complementary in identifying factors associated with higher crash severity. Cycling provides numerous benefits to individuals and to society but the burden of road traffic injuries and fatalities is disproportionately sustained by cyclists. Without awareness of the contributory factors of cyclist death and injury, the capability to implement context-specific and appropriate measures is severely limited. In this paper, we investigated the effects of the characteristics related to the road, the environment, the vehicle involved, the driver, and the cyclist on severity of crashes involving cyclists analysing 72,363 crashes that occurred in Great Britain in the period 2016–2018. Both a machine learning method, as the Random Forest (RF), and an econometric model, as the Random Parameters Logit Model (RPLM), were implemented. Three different RF algorithms were performed, namely the traditional RF, the Weighted Subspace RF, and the Random Survival Forest. The latter demonstrated superior predictive performances both in terms of F-measure and G-mean. The main result of the Random Survival Forest is the variable importance that provides a ranked list of the predictors associated with the fatal and severe cyclist crashes. For fatal classification, 19 variables showed a normalized importance higher than 5% with the second involved vehicle manoeuvring and the gender of the driver of the second vehicle having the greatest predictive ability. For serious injury classification, 13 variables showed a normalized importance higher than 5% with the bike leaving the carriageway having the greatest normalized importance. Furthermore, each path from the root node to the leaf nodes has been retraced the way back generating 361 if-then rules with fatal crash as consequent and 349 if-then rules with serious injury crash as consequent. The RPLM showed significant unobserved heterogeneity in the data finding four normal distributed indicator variables with random parameters: cyclist age ≥ 75 (fatal prediction), cyclist gender male (fatal and serious prediction), and driver aged 55–64 (serious prediction). The model's McFadden Pseudo R2 is equal to 0.21, indicating a very good fit. Furthermore, to understand the magnitude of the effects and the contribution of each variable to injury severity probabilities the pseudo-elasticity was assessed, gaining valuable insights into the relative importance and influence of the variables. The RF and the RPLM resulted complementary in identifying several roadways, environmental, vehicle, driver, and cyclist-related factors associated with higher crash severity. Based on the identified contributory factors, safety countermeasures useful to develop strategies for making bike a safer and more friendly form of transport were recommended. [ABSTRACT FROM AUTHOR]

• The performance of on-line vis-NIR for estimating soil organic carbon was evaluated. • Predictions for three datasets single field, UK with spiking, and UK were compared. • Random forest (RF) regression and spiking technique were used. • The best results obtained with the spiked UK dataset. Accurate on-line visible and near infrared (vis-NIR) spectroscopy prediction of soil organic carbon (OC) is essential for food security and environmental management. This paper aims at using on-line vis-NIR spectra coupled with random forest (RF) modelling approach for the prediction of soil organic carbon (OC), comparing between single field (SF), non-spiked UK multiple-field (NSUK) and spiked UK multiple-field (SUK) calibration models. Fresh soil samples collected from 6 fields in the UK (including two target fields) were scanned with a fibre-type vis-NIR spectrophotometer (tec5 Technology for Spectroscopy, Germany), with a spectral range of 305–2200 nm. After dividing spectra into calibration and independent validation sets, RF was run on the calibration set to develop calibration models for OC for the three studied datasets. Results showed that the model prediction performance depends on the dataset used and varies between fields. Less accurate prediction performance was obtained for the on-line prediction compared to laboratory (samples scanned in the laboratory under non-mobile measurement) prediction, and for non-spiked models compared to spiked models. The best model performance in both laboratory and on-line predictions was obtained when samples from the SF were spiked into the UK samples, with coefficients of determination (R2) values of 0.80 to 0.84 and 0.74 to 0.75, root mean square error of prediction (RMSEP) values of 0.14% and 0.17 to 0.18%, and ratio of prediction deviation (RPD) values of 2.30 to 2.5 and 1.98 to 2.04, respectively. Therefore, these results suggest that RF modelling approach when coupled with spiking provides high prediction performance of OC under both non-mobile laboratory and on-line field scanning conditions. [ABSTRACT FROM AUTHOR]

• Injury severity in single-bicycle crashes (SBCs) is analyzed. • A machine learning-based method termed Ordered Forest is employed. • ORF performs better than the traditional random forest in terms of predictions. • Factors associated with the injury severity in SBCs are investigated. • Practical guidance for the implementation of ORF is presented. This paper investigates the injury severity of cyclists in single-bicycle crashes (SBCs) in the UK. The data for analysis is constructed from the STATS19 road traffic casualty database, covering the period of 2016–2019. A machine learning-based ordered choice model termed Ordered Forest (ORF) is used. In our empirical analysis, ORF is found to produce more accurate class predictions of the SBC injury severity than the traditional random forest algorithm. Moreover, the factors associated with the injury severity are revealed, including the time and location of occurrence, the age of cyclists, roadway conditions, and crash-related factors. Specifically, old cyclists are more likely to be seriously injured in SBCs. Rural areas, higher speed limits, run-off crashes, and hitting objects are also related to an increased probability of serious injuries. While SBCs occurring at junctions, and/or during peak hours (i.e., 6:30–9:30 and 16:00–19:00) are less severe. To achieve the ambition of a step change in cycling and walking put forward by the UK Department for Transport, SBCs deserve more public attention. Lastly, regarding the implementation of ORF in crash injury severity analysis, we provide some practical guidance based on a series of simulation experiments. [ABSTRACT FROM AUTHOR]

Outbreaks of the COVID-19 pandemic caused by the SARS-CoV-2 infection that started in Wuhan, China, have quickly spread worldwide. The current situation has contributed to a dynamic rate of hospital admissions. Global efforts by Artificial Intelligence (AI) and Machine Learning (ML) communities to develop solutions to assist COVID-19-related research have escalated ever since. However, despite overwhelming efforts from the AI and ML community, many machine learning-based AI systems have been designed as black boxes. This paper proposes a model that utilizes Formal Concept Analysis (FCA) to explain a machine learning technique called Long-short Term Memory (LSTM) on a dataset of hospital admissions due to COVID-19 in the United Kingdom. This paper intends to increase the transparency of decision-making in the era of ML by using the proposed LSTM-FCA explainable model. Both LSTM and FCA are able to evaluate the data and explain the model to make the results more understandable and interpretable. The results and discussions are helpful and may lead to new research to optimize the use of ML in various real-world applications and to contain the disease. [ABSTRACT FROM AUTHOR]

The evolution towards Smart Grids (SGs) represents an important opportunity for modernization of the energy industry. It is characterized by a bidirectional flow of information and energy between consumers and suppliers. However, the rapid increase of energy demands in residential areas is becoming a challenging problem. In order to address this issue, Demand-Side Management (DSM) has proven to be an effective solution. In this paper, we propose LOSISH, a price-based Demand Response (DR) system for load scheduling in residential Smart Homes (SHs) that achieves a trade-off between electricity payments and consumer's discomfort. Our proposed system considers Renewable Energy Sources (RESs), Battery Energy Storage System (BESS) and Plug-in Electric Vehicle (PEV). We formulate our scheduling as a constrained optimization problem and we propose a new hybrid algorithm to solve it. The latter combines two well known heuristic algorithms: Particle Swarm Optimization (PSO) and Binary Particle Swarm Optimization (BPSO). Moreover, we propose a new clustering algorithm based on Machine Learning (ML) to extract consumer's preferences from a real dataset that contains the historical consumption patterns of his smart appliances. We test our approach on real data traces obtained from a SH and we set up an experiment to evaluate our algorithm on a Raspberry Pi and measure its energy consumption. To prove the effectiveness of our approach, we compare our results with another approach from the literature in terms of electricity bill, Peak-to-Average Ratio (PAR), energy consumption, and execution time. Numerical results show that LOSISH outperforms the other approach in terms of electricity bill (up to 52.92% cheaper), PAR (up to 44% decrease in peak demands), energy consumption (up to 69.44% less consumption), and execution time (up to 63.15% faster). • The load scheduling problem in residential smart homes is studied. • A constrained problem is formulated to make a trade-off between costs and comfort. • A hybrid heuristic algorithm and a Machine Learning clustering model are proposed. • Real data collected from a smart home in the UK was used to validate our approach. • Simulation and experimental results demonstrate the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

Figure. Graphical abstract (LCS = low-cost sensors; Ref = reference; AQ = air quality; Met = meteorology; GAM = generalised additive model; RF = random forest; QRM = quantile regression model; NO 2 = nitrogen dioxide). Air quality is monitored before intervention, then the intervention is implemented, and finally air quality is monitored again, which is analysed and modelled employing different machine learning approaches. [Display omitted] In this paper, the impact of smart traffic interventions on air quality was assessed in Thatcham, West Berkshire, UK. The intervention linked NO 2 levels with the cycle time of the traffic lights. When NO 2 levels exceeded a certain threshold, the strategy was triggered, which reduced the traffic congestion by turning the traffic lights green. Eight Earthsense Zephyrs air quality sensors and nine inductive-loop traffic detectors were installed in Thatcham to simultaneously monitor the air quality and traffic flows, respectively. Compared to the pre-intervention period, the observed NO 2 concentrations decreased in June, July and August and increased in September 2021, however, this does not reveal the true effect of smart traffic intervention. Using the observed data on the days with- and without-exceedances, we developed two machine learning models to predict the Business-as-usual (BAU) air quality level, i.e., a generalised additive model for average concentration and a quantile regression model for peak concentration. Our results demonstrated that average predicted concentrations (BAU) were lower than the observed concentrations (with intervention) by 12.45 %. However, we found that peak concentrations decreased by 20.54 %. [ABSTRACT FROM AUTHOR]

• A transfer learning architecture is proposed to improve the transferability of traffic prediction models. • A novel insight is provided to deal with data insufficient problems in traffic prediction. • The data selection criteria during model transfer are tested and discussed. Short-term traffic flow forecasting is a key element in Intelligent Transport Systems (ITS) to provide proactive traffic state information to road network operators. A variety of methods to predict traffic variables in the short-term can be found in the literature, ranging from time-series algorithms, machine learning tools and deep learning methods to a selective hybrid of these approaches. Despite the advances in prediction techniques, a challenging problem that affects the application of such methods in the real world is the prevalence of insufficient data across an entire network. It is rare that extensive historical training data required for model training are available for all the links in a city. In order to address this data insufficiency problem, this paper applies transfer learning techniques to machine learning methods in short-term traffic prediction. All the traffic data used in this paper were collected from Highways England road networks in the UK. The results show that through improving the transferability of machine learning-based models, the computational burden due to the model training process can be significantly reduced and the prediction accuracy under data deficient scenarios can be improved for one-step ahead prediction. However, the prediction accuracy gradually decreases in multi-step ahead prediction. It is also found that the accuracy of the proposed hybrid method is highly dependent upon consistency between datasets but less dependent on geographical attributes of links. [ABSTRACT FROM AUTHOR]

• Both spatial and temporal transferability of real-time crash risk models are assessed. • Wasserstein Generative Adversarial Network (WGAN) is introduced to address the extreme data imbalance issue. • Deep Neural Networks (DNN) and Transfer Learning (TL) are applied for predicting crashes. • Transfer learning improves transferability where direct transfer is not feasible. Real-time crash prediction is a heavily studied area given their potential applications in proactive traffic safety management in which a plethora of statistical and machine learning (ML) models have been developed to predict traffic crashes in real-time. However, one of the fundamental issues relating to the application of these models is spatio-temporal transferability. The present paper attempts to address this gap of knowledge by combining Generative Adversarial Network (GAN) and transfer learning to examine the transferability of real-time crash prediction models under an extremely imbalanced data setting. Initially, a baseline model was developed using Deep Neural Network (DNN) with crash and microscopic traffic data collected from M1 Motorway in the UK in 2017. The dataset utilised in the baseline model is naturally imbalanced with 257 crash cases and 16,359,163 non-crash cases. To overcome data imbalance issue, Wasserstein GAN (WGAN) was utilised to generate synthetic crash data. Non-crash data were randomly undersampled due to computational limitations. The calibrated model was then applied to predict traffic crashes for five other datasets obtained from M1 (2018), M4 (2017 & 2018 separately) and M6 Motorway (2017 & 2018 separately) by using transfer learning. Model transferability was compared with standalone models and direct transfer from the baseline model. The study revealed that direct transfer is not feasible. However, models become transferable temporally, spatially, and spatio-temporally if transfer learning is applied. The predictability of the transferred models outperformed existing studies by achieving high Area Under Curve (AUC) values ranging between 0.69 and 0.95. The best transferred model can predict nearly 95% crashes with only a 5% false alarm rate by tuning thresholds. Furthermore, the performances of transferred models are on par with or better than the standalone model. The findings of this study proves that transfer learning can improve model transferability under extremely imbalanced settings which helps traffic engineers in developing highly transferable models in future. [ABSTRACT FROM AUTHOR]

By constructing digital twins (DT) of an integrated energy system (IES), one can benefit from DT's predictive capabilities to improve coordinations among various energy converters, hence enhancing energy efficiency, cost savings and carbon emission reduction. This paper is motivated by the fact that practical IESs suffer from multiple uncertainty sources, and complicated surrounding environment. To address this problem, a novel DT-based day-ahead scheduling method is proposed. The physical IES is modelled as a multi-vector energy system in its virtual space that interacts with the physical IES to manipulate its operations. A deep neural network is trained to make statistical cost-saving scheduling by learning from both historical forecasting errors and day-ahead forecasts. Case studies of IESs show that the proposed DT-based method is able to reduce the operating cost of IES by 63.5%, comparing to the existing forecast-based scheduling methods. It is also found that both electric vehicles and thermal energy storages play proactive roles in the proposed method, highlighting their importance in future energy system integration and decarbonisation. • A deep learning and digital twins (DT) based scheduling method is proposed. • Physical IES performance is addressed by its virtual replica's predictive capability. • A machine learning method is proposed to address multiple uncertainty challenges. • Historical U.K. datasets are used for performance evaluations. • Results demonstrate the Digital Twin model reduces the extra operating cost by 63.5%. [ABSTRACT FROM AUTHOR]

The COVID-19 pandemic has provided psych challenges for many in society. One such challenge is the anxiety that is created in many people faced with the risk of death from the disease. Another issue is understanding how individuals cope psychologically with the threat of death from the disease. In this study we examine the manifestation of death anxiety and various coping mechanisms through the lens of terror management theory (TMT) and online platforms. We take a novel approach to testing the theory using big data analytics and machine learning, focusing on the user-generated content of Twitter users. Based on a sample of all tweets in the UK mentioning COVID-19 terms over a 5-month period, we evaluate dictionary mentions of anxiety and death, and various TMT defense mechanisms, and calculate the pattern of latent death anxiety or 'terror' states of Twitter users via Hidden Markov Models. The research identifies four online 'terror' states, with high death and anxiety mentions during the peak of the pandemic. Further we examine various TMT defense mechanisms that have been proposed in the literature for coping with death anxiety and find that online social connection, achievement and religion all play important roles in improving the model and explaining movement between states. The paper concludes with various implications of the study for future research and practice. • Extends terror management theory in an online social media context during the pandemic. • Examines death anxiety and online coping mechanisms from terror management theory. • Combines text analytics and hidden Markov models using a big data set. • Results support online social connection, achievement and religion coping mechanisms. • Contributes to the understanding of online individual behavior during pandemics. [ABSTRACT FROM AUTHOR]

Increasing the accuracy and intelligence of short-term load forecasting system can improve modern power systems management and economic power generation. In recent decades, the optimized machine learning methods have been widely used in load forecasting problems because of their predictability with higher accuracy and robustness. However, most related researches only use evolutionary algorithms for parameters fine-tuning and ignore the evolutionary algorithm based decision-making support and the matching relation between the used evolutionary algorithm and machine learning method, which greatly limit the improvement of forecasting system. To dissolve the above issues, a data-driven evolutionary ensemble learning forecasting model is proposed in this paper. Firstly, a novel multimodal evolutionary algorithm based on comprehensive weighted vector angle and shift-based density estimation is proposed. Secondly, based on the proposed multimodal evolutionary algorithm, an intelligent decision-making support scheme including predictive performance evaluation, model properties analysis, structure and fusion strategy optimization, and optimal model preference selection is designed to improve the random vector functional link network based ensemble learning model and boost the forecasting accuracy. Thirdly, experimental studies on 15 test problems with up to 6000 decision variables are conducted to validate the excellent optimization ability of the proposed evolutionary algorithm. Finally, the proposed evolutionary ensemble learning method is compared with 6 other representative forecast methods on real-world short-term load forecasting datasets from Australia, Great Britain, and Norway. The experiment results verify the superiority and applicability of the proposed method. • A new multimodal evolutionary algorithm (MMOWVA) is being proposed. • MMOWVA is validated by comparison experiments on 15 multimodal problems. • An expandable intelligent scheme is offered to make decision-makers convenient. • An evolutionary ensemble learning model is designed to improve forecast accuracy. • Experiment results demonstrate a higher forecast accuracy of the proposed method. [ABSTRACT FROM AUTHOR]

Geographic variation in digital inequality manifests as a result of a range of demographic, attitudinal, behavioural and locational factors. To better understand this multidimensional geography, our paper develops a new geodemographic classification for the spatial extent of Great Britain. In this model, we integrate a range of new small area measures that are drawn from multiple new forms of data including consumer purchasing data, survey and open data sources. Our analytical approach innovatively provides an integration of machine learning into a small-area estimation technique to obtain Lower Super Output Area / Data Zone estimates of Internet use, alongside a range of online engagement and consumption measures. Following the collation of a range of input measures, we implemented a more standard geodemographic framework that utilises the unsupervised clustering algorithm k-means to produce a map of the multidimensional characteristics of digital inequality for Great Britain; creating the Internet User Classification (IUC). Our outputs provide a new and nuanced understanding of the contemporary salient characteristics of digital inequality in Great Britain, which we evaluate both internally and externally within the context of preparations for the 2021 UK Census of the Population, exploring the geodemographic patterns of Census test response rates and the prevalence to complete the survey online. Our innovative work illustrates the strength of a geodemographic approach in mapping spatial patterns of digital inequality, and through the presented application concerning Census response rates and characteristics we demonstrate how the IUC can be operationalised within such settings for local intervention or benchmarking. • Great Britain can be differentiated into ten groups of salient digital inclusion / exclusion. • Supervised machine learning can provide methodological enhancement to small area estimation. • Census response rates have a relationship to patterns of digital inclusion / exclusion. [ABSTRACT FROM AUTHOR]

• Proposed an intelligent diagnosis framework for skin lesion classification. • Designed a multi-class multi-level algorithm to enhance the accuracy. • Proposed improved techniques for noise removal from the images. • Deep learning and other machine learning approaches developed and compared. • So far, the best multi-class result (∼96.5% accuracy) achieved using deep learning. Skin diseases remain a major cause of disability worldwide and contribute approximately 1.79% of the global burden of disease measured in disability-adjusted life years. In the United Kingdom alone, 60% of the population suffer from skin diseases during their lifetime. In this paper, we propose an intelligent digital diagnosis scheme to improve the classification accuracy of multiple diseases. A Multi-Class Multi-Level (MCML) classification algorithm inspired by the "divide and conquer" rule is explored to address the research challenges. The MCML classification algorithm is implemented using traditional machine learning and advanced deep learning approaches. Improved techniques are proposed for noise removal in the traditional machine learning approach. The proposed algorithm is evaluated on 3672 classified images, collected from different sources and the diagnostic accuracy of 96.47% is achieved. To verify the performance of the proposed algorithm, its metrics are compared with the Multi-Class Single-Level classification algorithm which is the main algorithm used in most of the existing literature. The results also indicate that the MCML classification algorithm is capable of enhancing the classification performance of multiple skin lesions. [ABSTRACT FROM AUTHOR]

• A novel algorithm named LWGF (Local Warning integrated with Global Feature) was proposed for FAIMS data analysis. • The LWGF identified 26 clinical patients (6 uninfected, 20 E. coli -infected) with the best average AUC of 0.98, and recognition rate of 96.15%. • For FAIMS data analysis, we have proven that it is better to use dynamic spectra rather than several stable spectra. Infections have long been a thorny problem that severely threatened public health and resulted in tremendous economic losses worldwide. Current detection methods for wound infection do not fully meet the requirements of preventing and treating this disease. Therefore, people are looking for better alternatives, wherein FAIMS (Field Asymmetric Ion Mobility Spectrometry) technology, by virtue of its high sensitivity, rapid response and noninvasive operation, is a promising candidate. This paper aims to investigate the possibility of FAIMS technology in detecting wound infections quickly and accurately. For this purpose, we gathered an odor dataset of clinical wound samples with the employment of a FAIMS instrument, the Lonestar (Owlstone, UK) analyzer. To enhance detectability, we proposed a novel algorithm framework, i.e., Local Warning integrated with Global Feature (LWGF), which is verified on distinguishment between twenty patients with single or mixed infection of Escherichia coli (E. coli) and six wounded patients without infection. Experimental results showed that the LWGF successfully identified the patients with the best average AUC of 0.98, and the best recognition rate of 96.15%, which are much higher than the conventional methods. [ABSTRACT FROM AUTHOR]