Individual Drivers Research Articles

A hybrid deep learning approach for driver anomalous lane changing identification

Reliable knowledge of driving states is of great importance to ensure road safety. Anomaly detection in driving behavior means recognizing anomalous driving states as a direct result of either environmental or psychological factors. This paper provides an efficient anomaly recognition approach to identify anomalous lane-changing events in a personalized manner. The proposed framework includes three unsupervised algorithms. First, a Recurrent-Convolutional Autoencoder extracts the spatio-temporal characteristics from a high-dimensional naturalistic driving dataset. Second, in order to recognize anomalous lane-changing events of individual drivers, the extracted latent feature space is analyzed using Pauta criterion-based reconstruction loss analysis, as well as one-class Support Vector Machine. Last, t-Distributed Stochastic Neighbor Embedding is employed to visualize the latent space for better understanding and interpretability. Temporal anomalies of lane-changing events were analyzed by a personalized grey relational coefficient analysis, to represent robust similarities for individual drivers. Validation and calibration were performed with a natural driving study dataset collected from 50 drivers with 59,372 lane change events. The results showed heterogeneity in the pattern of abnormal lane changing behavior across the sample. At the same time, each driver exhibited heterogeneous anomalous behaviors in both temporal and spatial sequences. Without prior labels, the proposed model effectively captures personalized driving patterns and abnormal lane-changing events from high-dimensional time-series data. This unsupervised hybrid approach is a novel attempt to complete personalized anomalous lane-changing behaviors identification based on naturalistic driving data involving various traffic environments. Our approach enables the extraction of natural individual lane-changing behavior patterns and provides insights for the improvement of personalized driving behavior monitoring systems.

Impact of Bus Rapid Transit Construction and Infrastructure on Traffic Safety: A Case Study From Albuquerque, New Mexico

Bus rapid transit (BRT) systems are becoming increasingly popular, yet their interaction with traffic safety in the United States has not been fully explored. How do BRT systems affect traffic safety, specifically for vulnerable road users such as pedestrians? Albuquerque, New Mexico recently installed the Albuquerque Rapid Transit (ART) system, a BRT system running along Central Avenue, a main east/west corridor through the city. Using collision data and volume counts for three snapshots in time (before, during, and after construction of the BRT), we analyzed the safety outcomes (all collisions and fatal/serious injury collisions) for all road users and pedestrians. We compared outcomes from the ART corridor to outcomes on alternative routes, on control segments of Central Avenue (those that did not see ART construction), and across the city. Collision counts saw the strongest decreases from before to after on the ART corridor (−8.2%). These benefits were more marked for fatal and serious injury collisions, with those collisions decreasing by 64.9% on ART (compared with a 5.7% decrease on control segments). Although vehicle volumes decreased an average of 28.6% after construction on Central Avenue, per-vehicle rates of fatal and serious collisions still decreased 57.1%. These results indicate that ART made the corridor safer overall (fewer total collisions), and while the risk to individual drivers increased, an individual’s chance of being fatally or seriously injured decreased significantly (−57.1%). Findings suggest that BRT systems can play an important role in the pursuit of a safe, healthy, and efficient transportation system.

Assessment of forest cover loss and impacts on ecosystem services: Coupling of remote sensing data and people's perception in the dry deciduous forest of West Bengal, India

Forest loss is one of the major environmental issues and is threatening livelihoods across the world. Understanding the drivers of forest cover loss in the different forest cover densities and its impacts on ecosystem services are relevant for sustainable management of the forest ecosystem. Extensive studies exploring topics related to spatial-temporal deforestation, little is known regarding quantitative forest loss by individual disturbance driver and people's perceptions of forest loss's impacts on provisioning ecosystem services. The present study focused on trends and patterns of forest loss by forest disturbance drivers in the different forest cover densities and its impacts on provisioning ecosystem services integrated with remote sensing data and local people's perceptions in the dry deciduous forest of West Bengal. This study employs stratified probability sampling approach using time series Landsat data on Google Earth Engine platform to quantify forest loss between 2006 and 2020 in the different pre-disturbance forest cover densities by forest disturbance drivers. Structured household interviews were used to examine local people's perceptions about forest cover loss and its impacts on ecosystem services. We found that 68.463% of forest area is cleared by disturbance drivers between 2006 and 2020. Annual rate of forest clearing is increasing trend and dominating in dense forests, followed by medium forests and open forests. Logging and agricultural clearing was the most dominant drivers over space and time. Overall 73.92% of households perceived decline in the availability of provisioning ecosystem services. Wild foods and medicinal plants were the most affected ecosystem services followed by livestock feed, construction materials and fuel wood. This study may help to the regional assessment of gross forest loss by individual disturbance driver, and its impacts on provisioning ecosystem services for supporting decision making to sustainable forest management.

‘Edible seaweeds’ as an alternative to animal-based proteins in the UK: Identifying product beliefs and consumer traits as drivers of consumer acceptability for macroalgae

Edible macroalgae (i.e., ‘seaweeds’) are a nutritious and sustainable alternative to animal-based proteins. However, consumption of seaweeds in Western countries remains low, and little is known about individual drivers of acceptance. The aim of this study was to further explore the consumer acceptability of seaweed-based food products in the UK. In an online study (N = 476), participants were presented with a general description of edible seaweeds, and descriptions of seaweed-based food products (e.g., ‘seaweed burger’). Participants were asked to rate beliefs about product attributes, and reported acceptance in terms of liking, willingness to try, willingness to buy, and readiness to adopt as a meat alternative. It was predicted that positive beliefs about seaweed-based products would be significantly associated with greater acceptance, and that seaweed-based products would be more favourable than a general description of seaweeds. Supporting study hypotheses, structural equation modelling showed that positive beliefs about taste/ edibility and familiarity significantly predicted acceptance (p <.01). Taste/ edibility was higher for seaweed-based products compared to a general description of seaweeds (p <.001), and perceiving foods to be tasty and familiar mediated the negative effect of food neophobia on consumer acceptance (p <.05). Other product beliefs – including cost, healthiness, and sustainability – were relatively poor predictors of acceptance (p >.05). These results support the consumer acceptance of seaweeds, and identify scope for utilising specific attributes of seaweeds (as drivers of acceptance) in future product development.

High-resolution mapping of wildfire drivers in California based on machine learning

Wildfires are important natural disturbances of ecosystems; however, they threaten the sustainability of ecosystems, climate and humans worldwide. It is vital to quantify and map the controlling drivers of wildfires for effective wildfire prediction and risk management. However, high-resolution mapping of wildfire drivers remains challenging. Here we established machine-learning (Random Forests) models using 23 climate and land surface variables as model inputs to reconstruct the spatial variability and seasonality of wildfire occurrence and extent in California. The importance of individual drivers was then quantified based on the Shapley value method. Thus, we provided spatially resolved maps of wildfire drivers at high resolutions up to 0.004° × 0.004°. The results indicated that precipitation and soil moisture are the major drivers dominating 37% of the total burnt area for large and extreme wildfires in summer and 63% in autumn, while elevation plays a major role for 15–58% of burnt areas in small wildfires in all seasons. Winds are also an important contributor to summer wildfires, accounting for 41% of large and extreme burnt areas. This study enhanced our knowledge of spatial variability of wildfire drivers across diverse landscapes in a fine-scale mapping, providing valuable perspectives and case studies for other regions of the world with frequently occurred wildfire.

Analyzing individual drivers of global changes promotes inaccurate long-term policies in deforestation hotspots: The case of Gran Chaco

In the context of the global climate and biodiversity crises, forecasting the effectiveness of Protected Areas (PAs) and forest management to conserve biodiversity in the long-term is a high priority, especially in threatened environments. By combining distribution models and conservation planning protocols, we analyzed the effect of global climate and agriculture-linked activities in the long-term conservation opportunities of one most threatened deforestation hotspots: the South American Gran Chaco. We showed that assessing the effects of each driver of global change individually, promotes inaccurate long-term policies in deforestation hotspots. Our future scenarios indicated a low impact of climate change on the species distributions when it was analyzed individually. However, its effects were strongly exacerbated when both drivers of threat were combined in the same analyses, strongly diminishing conservation opportunities in the region: more than 50% of the remaining species' distribution and hotspot areas could be lost in the near future. In this dramatic context, we identified important opportunities to improve the level of long-term protection by increasing at least 5.6% the protection coverage and placing PAs strategically. It is imperative policymakers promote policies to generate a long-term improvement of conservation areas that are resilient to both threats as soon as possible for these threatened environments.

THE INFLUENTIAL FACTORS OF WORK PRODUCTIVITY DURING COVID-19: THE EFFECT OF DIGITAL SKILL, COLLABORATION, AND ORGANIZATIONAL SUPPORT

For controlling the negative impact of Covid-19 outbreak, the Indonesia government instructs the citizen to work from home (WFH). Unfortunately, the WFH productivity is still questioned. This study is an attempt to elaborate work productivity and the effect of digital skill as individual driver, digital collaboration as group driver, and perceived organizational support or POS as organizational driver. For supporting the arguments, this article has conducted a quantitative study which involved 824 working citizens as the respondents from all over Indonesia. The respondents are mostly the first timers of working from home. They were working for governmental institution or business organizations from various industries in Indonesia. Collected data was structured and calculated by PLS structural equation modelling) and SmartPLS application version 3. This study concludes that digital skill and POS affect directly on work productivity, but digital collaboration impacts indirectly. Digital skill becomes a mediator in linkage between digital collaboration and work productivity

Capability development in hybrid organizations: enhancing military logistics with additive manufacturing

Additive Manufacturing (AM) has the potential to revolutionize key aspects of Military Logistics and partnerships between governmental and industrial organizations. Extreme outsourcing of key capabilities has created complex and deep hybrid organizations between armed forces and the private sector. In this study, the internal and external effects and requirements of Additive Manufacturing in the context of the hybrid organization of The Finnish Defence Forces (FDF) and its strategic partner in Maintenance, Repair, Overhaul (MRO), and Millog Oy were studied. First, with a literature review, we sought to link the capability development processes and the change drivers within them in both military and commercial contexts. Then, we utilized an existing, structured capability model used by the FDF (DOTMLPFI) and its individual change drivers to form an initial concept of AM as a part of the hybrid organization in question. The initial concept shows that AM can increase the performance of the commercially backed Military Logistic System by mitigating the risks of spare parts shortage in case of supply line disturbances and by facilitating localized spare parts production. However, the different primary goals of the military and commercial organizations and the contractual base of the hybrid organization impose constraints on the capability development process. Administrative decision-making across the organizations and the conflict between maximizing military and commercial potential are the key challenges in maintaining joint-capability systems of hybrid organizations.

Predictive Value of the Esterman Visual Field Test on the Outcome of the On-Road Driving Test.

PurposeAs the prevalence of age-related visual field disorders and the number of older drivers are rising, clear criteria on visual field requirements for driving are important. This article explores the predictive value of the Esterman visual field in relation to the outcome of an on-road driving test.MethodsA retrospective chart review was performed for driver's license applicants who, based on their visual field, performed an on-road driving test. Cases (N = 101) with a failed on-road driving test were matched with 101 controls with a passed outcome. The Esterman visual field was divided in regions, and the number of points missed per region was counted. Logistic regression models and receiver operating characteristic (ROC) curves were computed for each region.ResultsMost regions presented a significantly increased odds for failing the driving test when more points were missed. The odds ratio for the whole visual field was 2.52 (95% confidence interval, 1.53–4.14, P < 0.001) for all the participants. However, ROC curves failed to reveal distinct fail–pass criteria based on the number of points missed, as revealed by a large amount of overlap between cases and controls.ConclusionsThese findings confirm the relation between visual field damage and impaired driving performance. However, the Esterman visual field results were not conclusive for predicting the driving performance of the individual driver with visual field defects.Translational RelevanceIn our group of participants, the number of on-road driving tests cannot be further reduced by a more detailed definition of fail–pass criteria, based on the Esterman visual field test.

Sustainable Circular Economy in the Wood Construction Industry: A Business Opportunity Perspective

This study aims to explore the main drivers of recognizing a circular economy business opportunity. In this multiple case study with six entrepreneurs from the wood construction industry is a relevant part of the circular economy because its essential component is a sustainable and renewable material (trees/wood) that reduces carbon emissions. This study contributes to previous research by showing that six traditional drivers influence the entrepreneur’s recognition of business opportunities: environmental conditions, social capital, prior knowledge, systematic search, cognition and entrepreneurial alertness. Moreover, they are intertwined in the complex and processual phenomenon known as business opportunity. Furthermore, for the drivers of sustainable opportunity recognition, prior knowledge on environment and altruism are critical. The recognition of opportunity is not solely based on any of the individual drivers; rather, it relies mostly on continuous interaction with the network when the entrepreneur is creating a business opportunity.

Theory of planned behaviour, psychological stressors and intention to avoid violating traffic rules: A Multi-Level modelling analysis

This study applies a multilevel model approach to test the predictive effect of the theory of planned behaviour (TPB) plus moral norm, past behaviour and crash history to account for intentions to avoid traffic violations within the context of commuting to or from work. This study also extended the theory by adding psychological stressors of perceived stress and work-family conflict. In this study, we systematically tested the direct and mediated models. A sample of Malaysian drivers (N = 482; 44.6% were men and 55.4% women), with the average age of 36.7 years (SD = 10.0) was surveyed. The self-report questionnaire contained multiple observations nested within individual drivers, with respect to 3 different driving violations. As predicted, multi-level modelling showed that within-person predictor variables of all the TPB components emerged as independent predictors, with injunctive norm being the most predictive variable, followed by self-efficacy and cognitive attitude. Intention was also predicted by moral norms and crash history. Between-person variables, gender, commuting hours and work-family conflict also had direct influences on intention (i.e., men and those with longer commuting hours and higher work-family conflict reported lower intentions). Substantial support for the mediation model was found, confirming that stressors indirectly influence intentions through effects on the social cognitive components. The implications of the linkage between the social cognition and stressors for developing potential broader interventions focusing on multiple violation behaviours and designing appropriate safety policies to reduce commuting crashes are discussed.

Influence of the Placement Accuracy of the Brushless DC Motor Hall Sensor on Inverter Transistor Losses

Low-power BLDC motors are often and willingly used in many drive devices due to their functional advantages. They are also used in advanced positioning systems, where their good dynamic performance parameters are used. The control systems use shaft position sensors mounted on motors, the structure of which is based on magnetic elements and Hall sensors. The aim of this article was to investigate the influence of the BLDC motor quality on the correct operation of the control semiconductor system. The article presents the effect of BLDC motor shaft observation system’s inaccuracies on the friction and current amplitudes of individual inverter keys. Waveforms of the controller phase currents are considered and recorded on a test bench that allows precise sensor position changes. In addition, the effect of sensor misalignment on power losses in individual inverter transistors is investigated. The article shows a significant influence of the motor shaft observation system’s assembly accuracy on the current amplitudes of individual driver transistors and their power losses, which makes it necessary to consider these parameters when constructing power electronic systems.

Large-scale scenarios of electric vehicle charging with a data-driven model of control

Transportation electrification is forecast to bring millions of new electric vehicles to roads worldwide this decade. Planning to support those vehicles depends on detailed scenarios of their electricity demand in both uncontrolled and controlled or smart charging scenarios. In this paper, we present a novel modeling approach to enable rapid generation of demand estimates that represent the impact of controlled charging for large-scale scenarios with millions of individual drivers. To model the effect of load modulation control on aggregate charging profiles, we propose a novel machine learning approach that replaces traditional optimization approaches. We demonstrate its performance modeling workplace charging control under a range of electricity rate schedules, achieving small errors (2.5%–4.5%) while accelerating computations by more than 4000 times. To generate the uncontrolled charging demand for scenarios with residential, workplace, and public charging we use statistical representations of a large data set of real charging sessions. We demonstrate the methodology by generating diverse sets of scenarios for California's charging demand in 2030 which consider multiple charging segments and controls, each run locally in under 50 s. We further demonstrate support for rate design by modeling the large-scale impact of a new, custom rate schedule for workplace charging.

Personalized driving assistance algorithms: Case study of federated learning based forward collision warning

Current designs of advanced driving assistance systems (ADAS) mainly developed uniform collision warning algorithms, which ignore the heterogeneity of driving behaviors, thus lead to low drivers’ trust in. To address this issue, developing personalized driving assistance algorithms is a promising approach. However, current personalization systems were mainly implemented through manually adjusting warning trigger thresholds, which would be less feasible for overall drivers as certain domain expertise is required to set personal thresholds accurately. Other personalization techniques exploited individual drivers’ data to build personalized models. Such approach could learn personal behavior but requires impractical large-scale individual data collections. To fill up the gaps, self-adaptive algorithms for personalized forward collision warning (FCW) based on federated learning were proposed in this study. A baseline model was developed by long short-term memory (LSTM) for FCW. Federated learning framework was then introduced to collect knowledge from multiple drivers with privacy preserving. Specifically, a general cloud server model was trained by collecting updated parameters from individual vehicle server models rather than collecting raw data. Besides, a driver-specific batch normalization (BN) layer was added into each vehicle server model to address the heterogeneity of driving behaviors. Experiments show empirically that the proposed federated-based personalized models with the BN layer showed to have the best performance. The average modeling accuracy has reached 84.88% and the performance is comparable to conventional total data collection training approach, where the additional BN layer could increase the accuracy by 3.48%. Finally, applications of the proposed framework and its further investigations have been discussed.

Association between biomarkers and clinical outcomes of lenvatinib + pembrolizumab in advanced renal cell carcinoma (RCC): Results from Study 111/KEYNOTE-146.

375 Background: In the Study 111/KEYNOTE-146 trial (NCT02501096; N=147), lenvatinib (lenva) + pembrolizumab (pembro) showed encouraging antitumor activity and a manageable safety profile in treatment-naive (n=23) or previously treated metastatic RCC (n=105, previously treated with immune checkpoint inhibitor [ICI]; n=19, previously treated ICI naive); 145 had clear cell RCC and 2 had non-clear cell RCC. In this exploratory analysis, we evaluated the association between clinical outcomes and gene expression signatures and DNA variants for individual RCC-specific driver genes of interest based on published reports. Methods: Patients (pts) with metastatic RCC were treated with lenva 20 mg orally once daily + pembro 200 mg intravenously once every 3 weeks. The analysis population included pts with treatment-naive (n=10) and ICI pretreated (n=70) disease with evaluable RNA-sequencing data for the 18-gene T-cell–inflamed gene expression profile (TcellinfGEP) and for 11 other signatures (angiogenesis; glycolysis; gMDSC; hypoxia; mMDSC; MVD; MYC; proliferation; RAS; stroma/EMT/TGFβ; WNT) and whole exome sequencing (WES) data for DNA variants for individual genes ( VHL, PBRM1, BAP1, and SETD2). Specimens were collected prior to the start of treatment. The associations between each signature score and ORR and PFS per immune-related RECIST were evaluated using logistic regression and Cox proportional hazards, respectively. One-sided P values for TcellinfGEP (hypothesized positive association) and two-sided P values for all other signatures (no hypothesized association) were adjusted for multiplicity using the Hochberg step-up procedure; significance was prespecified at α=0.05. The association between DNA variants for individual genes and ORR was evaluated descriptively. Clinical data cutoff was August 18, 2020. Results: Of 147 treated pts, RNA sequencing and WES data were available for 80 (54%) and 60 (41%), respectively. TcellinfGEP was not associated with ORR ( P=0.827) or PFS ( P=0.741), nor were the other 11 signatures before or after adjustment for TcellinfGEP. ORR for DNA variants reported in the table. Conclusions: In this exploratory analysis of pts with metastatic RCC enrolled in Study 111/KEYNOTE-146 treated with lenva + pembro, responses were observed regardless of biomarker status. There were no statistically significant associations between gene signatures and clinical outcomes. Clinical benefit was observed regardless of VHL, PBRM1, BAP1, or SETD2 mutation status. Analyses in larger randomized datasets will provide additional information on the role of biomarkers in RCC. Clinical trial information: NCT02501096. [Table: see text]

Survey Based Road Safety Analysis to Understand the Perspective of the General Public

The number of road traffic deaths continues to climb, reaching 1.35 million deaths globally in 2016 and road traffic injury is now the leading cause of death for children and young adults aged 5–29 years (WHO, 2018). Road safety is officially recognized as a global safety issue and it is also included in the Sustainable Development Goals. This is a questionnaire based analysis, which aims to understand the outlook of the general public regarding the current road safety system and to understand the level of awareness regarding the same. It also aims to compare the perspective of drivers and non-drivers towards traffic rules. Majority of the participants were from the age group of 15-30 years as they are most directly affected by road accidents. Our objective is to understand the shortcomings in the current road safety system through the perspective of the general public, which can then help in implementing a more efficient system in order to decrease the number of road accidents. Inadequate traffic system also significantly affects the economic growth of the region. According to the World Road Statistics, 2018, India ranks 1st in the number of road accident deaths hence there is an urgent need to understand the root cause for such a high number of accidents and to find solutions for this issue. Our results show that there is a need to implement stricter rules for road safety and this perspective of the general public is not affected by the status of the individual (driver, non-driver). It also suggests the need for increasing the level of licence tests conducted. Increasing awareness at an early stage among the people regarding the seriousness of following traffic rules can help prevent road accidents at the baseline level

Autonomous Vehicle Path Planning Based on Driver Characteristics Identification and Improved Artificial Potential Field

Different driving styles should be considered in path planning for autonomous vehicles that are travelling alongside other traditional vehicles in the same traffic scene. Based on the drivers’ characteristics and artificial potential field (APF), an improved local path planning algorithm is proposed in this paper. A large amount of driver data are collected through tests and classified by the K-means algorithm. A Keras neural network model is trained by using the above data. APF is combined with driver characteristic identification. The distances between the vehicle and obstacle are normalized. The repulsive potential field functions are designed according to different driver characteristics and road boundaries. The designed local path planning method can adapt to different surrounding manual driving vehicles. The proposed human-like decision path planning method is compared with the traditional APF planning method. Simulation tests of an individual driver and various drivers with different characteristics in overtaking scenes are carried out. The simulation results show that the curves of human-like decision-making path planning method are more reasonable than those of the traditional APF path planning method; the proposed method can carry out more effective path planning for autonomous vehicles according to the different driving styles of surrounding manual vehicles.

Survey Based Road Safety Analysis to Understand the Perspective of the General Public

The number of road traffic deaths continues to climb, reaching 1.35 million deaths globally in 2016 and road traffic injury is now the leading cause of death for children and young adults aged 5–29 years (WHO, 2018). Road safety is officially recognized as a global safety issue and it is also included in the Sustainable Development Goals. This is a questionnaire based analysis, which aims to understand the outlook of the general public regarding the current road safety system and to understand the level of awareness regarding the same. It also aims to compare the perspective of drivers and non-drivers towards traffic rules. Majority of the participants were from the age group of 15-30 years as they are most directly affected by road accidents. Our objective is to understand the shortcomings in the current road safety system through the perspective of the general public, which can then help in implementing a more efficient system in order to decrease the number of road accidents. Inadequate traffic system also significantly affects the economic growth of the region. According to the World Road Statistics, 2018, India ranks 1st in the number of road accident deaths hence there is an urgent need to understand the root cause for such a high number of accidents and to find solutions for this issue. Our results show that there is a need to implement stricter rules for road safety and this perspective of the general public is not affected by the status of the individual (driver, non-driver). It also suggests the need for increasing the level of licence tests conducted. Increasing awareness at an early stage among the people regarding the seriousness of following traffic rules can help prevent road accidents at the baseline level.

Predicting individuals' car accident risk by trajectory, driving events, and geographical context

With the prevalence of GPS tracking technologiees, car insurance companies have started to adopt usage-based insurance policies, which adapt insurance premiums according to the customers' driving behavior. Although many risk models for assessing an individual driver's accident risk based on the history of driving trajectories, driving events, and exposure records exist, these models do not take the geographical context of the driven trajectories and driving events into account. This study explores the influence of enriching the existing purely driving-behavior-based feature set by multiple geographical context features for the task of differentiating between accident and accident-free drivers. Prediction performances of five machine learning classifiers—logistic regression, random forest, XGBoost, feed-forward neural networks (FFNN), and long short-term memory (LSTM) networks—were evaluated on the usage records of over 8000 vehicles in one year from Italy. The results show that the inclusion of geographical information such as weather, points of interest (POIs), and land use can increase the relative predictive performance in terms of AUC by up to 8%, among which land use is the most informative. For the data of this study, XGBoost generally yielded the best performance and made most use out of the geographical information, while logistic regression is only slightly outperformed by more complex models if the proposed geographical information is not available. LSTM did not outperform the other methods, possibly due to the small volume of training data available. The results outline the potential of including the geographical context in usage-based car insurance risk modeling to improve the accuracy, leading to fairer usage-based insurance policies.

Forecasting asylum-related migration flows with machine learning and data at scale

The sudden and unexpected migration flows that reached Europe during the so-called ‘refugee crisis’ of 2015–2016 left governments unprepared, exposing significant shortcomings in the field of migration forecasting. Forecasting asylum-related migration is indeed problematic. Migration is a complex system, drivers are composite, measurement incorporates uncertainty, and most migration theories are either under-specified or hardly actionable. As a result, approaches to forecasting generally focus on specific migration flows, and the results are often inconsistent and difficult to generalise. Here we present an adaptive machine learning algorithm that integrates administrative statistics and non-traditional data sources at scale to effectively forecast asylum-related migration flows. We focus on asylum applications lodged in countries of the European Union (EU) by nationals of all countries of origin worldwide, but the same approach can be applied in any context provided adequate migration or asylum data are available. Uniquely, our approach (a) monitors drivers in countries of origin and destination to detect early onset change; (b) models individual country-to-country migration flows separately and on moving time windows; (c) estimates the effects of individual drivers, including lagged effects; (d) delivers forecasts of asylum applications up to four weeks ahead; (e) assesses how patterns of drivers shift over time to describe the functioning and change of migration systems. Our approach draws on migration theory and modelling, international protection, and data science to deliver what is, to our knowledge, the first comprehensive system for forecasting asylum applications based on adaptive models and data at scale. Importantly, this approach can be extended to forecast other social processes.