Predicting Driver Fatigue in Monotonous Automated Driving with Explanation using GPBoost and SHAP

Zhou, Feng; Alsaid, Areen; Blommer, Mike; Curry, Reates; Swaminathan, Radhakrishnan; Kochhar, Dev S.; Talamonti, Walter; Tijerina, Louis

doi:10.1080/10447318.2021.1965774

Cited by 15 publications

(5 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The feature significance shown in the Results section for XGBoost, LightGBM, and GPBoost were obtained from SHAP (Shapley Additive eXplanations) values of terms for the test data set using the TreeExplainer method within the SHAP module ( https://shap.readthedocs.io/ ) in Python 3.7 [54] . Among the principal reasons for selecting GPBoost to implement mixed effects machine learning was its compatibility with SHAP for interpretation [50] , [74] . Feature importance can also be derived for the aforementioned ensemble decision tree methods by computing, for example, the number of times a feature is used to split trees, or the gain in score towards the objective function obtained by splitting trees based on a feature [75] , [76] .…”

Section: Methodsmentioning

confidence: 99%

Predicting COVID-19 disease severity from SARS-CoV-2 spike protein sequence by mixed effects machine learning

Sokhansanj

Rosen

2022

Computers in Biology and Medicine

View full text Add to dashboard Cite

Section: Methodsmentioning

confidence: 99%

Predicting COVID-19 disease severity from SARS-CoV-2 spike protein sequence by mixed effects machine learning

Sokhansanj

Rosen

2022

Computers in Biology and Medicine

View full text Add to dashboard Cite

“…We fitted this spatial ML model using the gpboost R package (Sigrist, 2023), which uses the LightGBM algorithm (Ke et al, 2017), and the structured spatial effects are modeled using Gaussian processes with a Matérn covariance kernel (Gelfand & Schliep, 2016). See Zhou et al (2022) for a recent example of using GPBoost with SHAP-based explanations to predict and interpret manual car driver fatigue or Sokhansanj and Rosen (2022) for predicting COVID-19 disease severity.…”

Section: Spatial ML Modelsmentioning

confidence: 99%

Evidence from interpretable machine learning to inform spatial management of Palau's tuna fisheries

Gilman,

Chaloupka

2024

Ecosphere

View full text Add to dashboard Cite

Static and dynamic area‐based management tools hold substantial potential to balance socioeconomic benefits derived from fisheries and costs from bycatch mortality of at‐risk species. Palau longline fisheries have high bycatch of at‐risk species including the olive ridley marine turtle and silky and blue sharks. This study analyzed a two decades‐long time series of observer and electronic monitoring datasets from the Palau distant‐water and locally‐based pelagic longline fisheries. An interpretable or explainable machine learning‐based modeling approach was used to derive spatially resolved species‐specific catch rate predictions. These models were conditioned on a suite of potentially informative environmental, bathymetric, ocean‐climate metric, vessel, monitoring system, and set‐specific operational predictors. Overall, there would be limited ecological tradeoffs from focusing fishing effort within primary catch rate hotspots for target bigeye and yellowfin tunas. Mean field prediction surfaces also defined catch rate hotspots for at‐risk species of silky and blue sharks, olive ridley turtle, and pelagic stingray, which did not overlap the hotspots for target species. The predicted target species hotspots, however, overlap olive ridley and pelagic stingray warmspots. Results also identify opportunities for temporally dynamic spatial management to control catch rates of target and bycatch species. Management of fishery operational predictors of fishing depth and soak duration present additional opportunities to balance catch rates of at‐risk bycatch and target species. A transition to employing fleetwide or vessel‐based output controls that effectively constrain the fishery would alter the spatial management strategy to focus on zones with the lowest ratio of at‐risk bycatch to commercial catch. Our findings support evidence‐informed evaluation of spatial management strategies and complementary measures to meet objectives for balancing socioeconomic benefits derived from target species catch with costs to threatened species.

show abstract

“…Therefore, researchers proposed algorithms to detect and predict the intentions of drivers to alert them of risky traffic situations and make decisions about whether to take over the driver's control of the vehicle [18,25,28]. Other studies also examined the methods to improve driver's behavior modeling and detection, such as eye gaze detection, fatigue detection, and other driving behavior detection for better monitoring performance [64,67,[70][71][72][73][74]. In terms of drivers' intentional disuse of the system, Meuller et al [43] proposed a driver monitoring system based on driver behavior tracking to detect deliberate disengagement and misuse of ADAS, the system then used attention reminders and proactive methods to keep drivers engaged in the driver-system interactions and maintain the system's functionalities.…”

Section: Adas Solutionsmentioning

confidence: 99%

Cause-and-Effect Analysis of ADAS: A Comparison Study between Literature Review and Complaint Data

Ayoub

Wang

et al. 2022

Proceedings of the 14th International Conference on Automotive User Interfaces and Interactive Vehicular Applications

Self Cite

View full text Add to dashboard Cite

Advanced driver assistance systems (ADAS) are designed to improve vehicle safety. However, it is difficult to achieve such benefits without understanding the causes and limitations of the current ADAS and their possible solutions. This study 1) investigated the limitations and solutions of ADAS through a literature review, 2) identified the causes and effects of ADAS through consumer complaints using natural language processing models, and 3) compared the major differences between the two. These two lines of research identified similar categories of ADAS causes, including human factors, environmental factors, and vehicle factors. However, academic research focused more on human factors of ADAS issues and proposed advanced algorithms to mitigate such issues while drivers complained more of vehicle factors of ADAS failures, which led to associated top consequences. The findings from these two sources tend to complement each other and provide important implications for the improvement of ADAS in the future.

show abstract

Predicting Driver Fatigue in Monotonous Automated Driving with Explanation using GPBoost and SHAP

Cited by 15 publications

References 43 publications

Predicting COVID-19 disease severity from SARS-CoV-2 spike protein sequence by mixed effects machine learning

Predicting COVID-19 disease severity from SARS-CoV-2 spike protein sequence by mixed effects machine learning

Evidence from interpretable machine learning to inform spatial management of Palau's tuna fisheries

Cause-and-Effect Analysis of ADAS: A Comparison Study between Literature Review and Complaint Data

Contact Info

Product

Resources

About