Time Prediction Research Articles

Gravitational wave signal denoising and merger time prediction with a deep neural network

Gravitational wave signal denoising and merger time prediction with a deep neural network

Combining non-targeted high resolution mass spectrometry with effect-directed analysis to identify contaminants of emerging concern in the field of ecotoxicology: A systematic quantitative literature review.

Methods for measuring environmental toxicity and identifying chemical toxicity drivers using non-targeted analysis (NTA) were reviewed in this systematic quantitative literature review. Effect-directed analysis (EDA) was used to assess sample toxicity and prioritise NTA sample analysis. The most common bioassays performed were estrogen, androgen and aryl hydrocarbon receptor assays, with many studies using test batteries. Across the 95 studies in this review, the toxicity could be explained (>75%) for eight studies, four studies had toxicity endpoints explained and unexplained, and 38 studies had unexplained (<75%) toxicity. The addition of NTA allowed for toxicity to be explained with a median of 47% for TOXnon-target studies and 34% for TOXtarget+non-target, far higher than the 13% median for TOXtarget studies within this review. The outcomes of identification were affected by method factors including sample extraction, chromatography method, data acquisition and data processing. Method factors with the biggest potential to introduce selection bias were sample extraction and chromatography technique. These factors were characterised by a high representation of reverse phase liquid chromatography contributing to the selective exclusion of polar, highly polar and ionic compounds from sample analysis. This reduces compound identification and excludes unknown chemical contaminants from analysis. Not all studies reported the explained toxicity contribution of identified compounds, however it was evident that many compound features could not be identified using NTA processing software. There were severe limitations for liquid chromatography data compared to gas chromatography data with insufficient spectral library databases for spectra matching. This bottleneck is partially solved through the rise in in silico and retention time prediction software. Future work, including increasing spectral databases for liquid chromatography, use of less biased chromatography and sample preparation techniques and inclusion of EDA-NTA into risk assessment frameworks, will allow for better toxicity assessment of emerging contaminants.

Continuous time and dynamic suicide attempt risk prediction with neural ordinary differential equations

Current clinician-based and automated risk assessment methods treat the risk of suicide-related behaviors (SRBs) as static, while in actual clinical practice, SRB risk fluctuates over time. Here, we develop two closely related model classes, Event-GRU-ODE and Event-GRU-Discretized, that can predict the dynamic risk of events as a continuous trajectory across future time points, even without new observations, while updating these estimates as new data become available. Models were trained and validated for SRB prediction using a large electronic health record database. Both models demonstrated high discrimination (e.g., Event-GRU-ODE AUROC = 0.93, AUPRC = 0.10, relative risk = 13.4 at 95% specificity in a low-prevalence [0.15%] general cohort with a 1.5-year prediction window). This work provides an initial step toward developing novel suicide prevention strategies based on dynamic changes in risk.

EXPRESS: Lead Time Prediction for Inventory Optimization with Machine Learning

EXPRESS: Lead Time Prediction for Inventory Optimization with Machine Learning

Applications and Advancements of Spaceborne InSAR in Landslide Monitoring and Susceptibility Mapping: A Systematic Review

Landslides pose significant threats to human safety and socio-economic development. In recent decades, interferometric synthetic aperture radar (InSAR) technology has emerged as a powerful tool for investigating landslides. This study systematically reviews the applications of spaceborne InSAR in landslide monitoring and susceptibility mapping over the past decade. We highlight advancements in key areas, including atmospheric delay correction, 3D landslide monitoring, failure time prediction, enhancements in spatial and temporal resolution, and integration with other technologies like the Global Navigation Satellite System (GNSS) and physical models. Additionally, we summarize various InSAR application strategies in landslide susceptibility mapping, identifying a gap between the static nature of most current studies and InSAR’s dynamic potential for capturing deformation velocity. Future research should integrate InSAR-derived factors with other dynamic variables like rainfall and soil moisture for dynamic susceptibility mapping and prediction. We also emphasize that further development of dynamic InSAR will require more efficient SAR data management and processing strategies.

An Investigation of Pre-stimulus EEG for Prediction of Driver Reaction Time.

Driver drowsiness significantly contributes to road accidents worldwide, and timely prediction of driver reaction time is crucial for developing effective advanced driver assistance systems. In this paper, we present an EEG-based prediction framework that investigates the impact of different pre-stimulus time windows, frequency band combinations, and channel groups on driver reaction time estimation using data from a 90-minute sustained attention driving task. Our systematic evaluation using a publicly available dataset of 24 drivers reveals that a 2-second pre-stimulus window yields the lowest prediction error. Notably, our proposed 1D Convolutional Neural Network (CNN) approach reduces the Mean Absolute Error (MAE) by nearly 30\% (from 0.51 sec to 0.36 sec for the alpha band) compared to classical machine learning models. Moreover, while individual frequency bands (e.g., alpha and theta) outperform combined band approaches, most spatial channel groups deliver similar performance to the full 32-channel configuration-with the exception of frontal channels. These improvements underscore the potential for real-world applications in reducing road accidents by enabling timely interventions based on predictive analytics.

Prediction of time averaged wall shear stress distribution in coronary arteries’ bifurcation varying in morphological features via deep learning

IntroductionUnderstanding the hemodynamics of blood circulation is crucial to reveal the processes contributing to stenosis and atherosclerosis development.MethodComputational fluid dynamics (CFD) facilitates this understanding by simulating blood flow patterns in coronary arteries. Nevertheless, applying CFD in fast-response scenarios presents challenge due to the high computational costs. To overcome this challenge, we integrate a deep learning (DL) method to improve efficiency and responsiveness. This study presents a DL approach for predicting Time-Averaged Wall Shear Stress (TAWSS) values in coronary arteries’ bifurcation.ResultsTo prepare the dataset, 1800 idealized models with varying morphological parameters are created. Afterward, we design a CNN-based U-net architecture to predict TAWSS by the point cloud of the geometries. Moreover, this architecture is implemented using TensorFlow 2.3.0. Our results indicate that the proposed algorithms can generate results in less than one second, showcasing their suitability for applications in terms of computational efficiency.DiscussionFurthermore, the DL-based predictions demonstrate strong agreement with results from CFD simulations, with a normalized mean absolute error of only 2.53% across various cases.

Neural network insights: explainable artificial intelligence (XAI) and hyper-parameter dynamics for precise travel time predictions

Urban transportation systems face increasing challenges due to population growth, urbanization, and dynamic mobility patterns. Accurate travel time prediction is critical for efficient traffic management, urban planning, and transportation infrastructure optimization. This study investigates travel time prediction in urban environments through a two-phase approach combining advanced Artificial Neural Network (ANN) architectures and interpretability techniques. In the first phase, we systematically analyze the impact of ANN architectural depth, width, and key hyper-parameters on predictive accuracy, highlighting the interplay between model configurations and performance. In the second phase, we employ SHapley Additive exPlanations (SHAP) to investigate hyper-parameter-feature interaction, and enhance model transparency. Additionally, an in-depth error and uncertainty analysis for the best-performing model provides insights into residual behavior, prediction intervals, and areas for refinement. The results demonstrate the robustness and adaptability of ANN models in capturing complex spatiotemporal dependencies, while SHAP-based analysis offers actionable insights into feature contributions. Ultimately, this study advances travel time prediction understanding and offers a comprehensive AI framework for urban mobility studies, facilitating the development of precise and adaptive transportation systems.

Prediction of open-pit mine truck travel time based on LSTM-TabTransformer

Truck travel time prediction is the basis for real-time optimal scheduling decision of open-pit trucks. Most prediction models for open-pit truck travel time mainly rely on a single machine learning algorithm or model to optimize the hyperparameters, which is difficult to accurately capture the composite characteristics of truck travel time data, affectting the accuracy of the prediction results. This paper proposed a travel time prediction method for open-pit trucks based on LSTM-TabTransformer. After eliminating the outliers in the truck travel time data set by applying the Paura criterion, the data set was non-dimensionalized, and the corresponding category data was processed by means of data diversion. Then, the self-attention mechanism of TabTransformer and gating mechanism of LSTM were used to capture the dynamic characteristics of the travel time variation rule at multiple levels. Finally, the dimensionality of the concatenated feature matrix was reduced by MLP, and the predicted truck travel time was output. Based on the truck travel time series data set collected from the example open-pit coal mine, the prediction experiment was carried out. The results show that the combined machine learning model LSTM-Tabtransformer proposed in this paper is suitable to process the complex truck travel time series data set and learn multiple features, which can reduce the volatility of relative absolute error (RAE) and relative square error (RSE) of travel time prediction, overcome the limitations of a single machine learning prediction model, effectively increase the sensitivity of learning features, and significantly improve the accuracy of truck travel time prediction in open-pit mines.

DEVELOPMENT OF DEEP LEARNING MULTI – LAYER PERCEPTRON MODEL USING NEGATIVE BINOMIAL REGRESSION APPROACH FOR WAITING TIME PREDICTION

Evidently queuing model has been very useful in identifying appropriate levels of staff, equipment, and beds along with in making decisions about resource allocation and the design of new services as well as Waiting Time (WT) estimation and predictions. However, the traditional Queuing Theory approach has been known not to be sufficient in real life applications because the methodology is limited, for instance, unrealistic assumptions of the time distribution it requires to do queuing analysis. Thus, the goal of this study is to develop deep learning multi – layer perceptron model using negative binomial regression approach for waiting time prediction and examine the existing used ML algorithms for WT prediction at different sample sizes of queuing and make an essential tool for responsive actions for any organization (i.e. healthcare centers) reporting long waiting times. The study sets to follow Monte Carlos Simulation process and utilization of real-life data. Sequel to the developed multi-layer poisson regression (MLP-PR) and following the shortcomings of poisson regression to account for overdispersion in count data as well as to regularize the problem of over-fitting in MLP approach, this pursues to introduce a Novel Multi-Layer Perceptron Negative Binomial Regression (MLP-NBR)for WT Prediction. The Novel model is expected to handle problem of overdispersion, autocorrelation and over-fitting more effectively compare to the MLP-PR.

A computational model for prediction of IR intensity and burn time of Magnesium-Teflon-Viton (MTV) based infrared (IR) decoy flare of various configurations

A computational model for prediction of IR intensity and burn time of Magnesium-Teflon-Viton (MTV) based infrared (IR) decoy flare of various configurations

Multi-aircraft attention-based model for perceptive arrival transit time prediction

Multi-aircraft attention-based model for perceptive arrival transit time prediction

Deep learning and finite element simulation of spatiotemporal multiscale fusion: Real time prediction of magnesium alloys corrosion covered with MAO coatings

Deep learning and finite element simulation of spatiotemporal multiscale fusion: Real time prediction of magnesium alloys corrosion covered with MAO coatings

Task migration with deadlines using machine learning-based dwell time prediction in vehicular micro clouds

Task migration with deadlines using machine learning-based dwell time prediction in vehicular micro clouds

Prediction of battery charging time and range detection in electric vehicle using machine learning algorithms

As a key pillar of smart transportation in smart city applications, Electric Vehicles (EVs) are becoming increasingly popular for their contribution in reducing greenhouse gas emissions. The lack of charging infrastructure and range detection is one of the most significant barriers to Electric Vehicle adoption. To address the problem of EV charging and range detection, employing Machine Learning (ML) algorithms to predict charging analysis, which is beneficial to drivers. ML algorithms in Electric Vehicle (EV) refers to the application of computational algorithms and statistical models that enable EVs and their supporting systems. To train the model own data set was created using hardware setup. This hardware setup consists of Battery, ESP Controller, Sensors, and Arduino IDE. Here two type of machine learning algorithms where we used labelled dataset to train the model or algorithms. one is the Support Vector Regression (SVR), other one is Random Forest Algorithm (RFA). It predicts charging patterns, recommend optimal charging times, and estimate charging times based on factors like battery health, current charge level, and environmental conditions. This leads to more efficient charging and better energy management and also going to predict the driving range of EVs based on performance metric like RMSE (Root Mean Squared Error), MAE (Mean Absolute Percentage Error), R Square, SMAPE (Symmetrical Mean Absolute Percentage Error). This information aids drivers in planning routes and making informed decisions. ML in the Electric Vehicle industry transforms data into actionable insights that enhance vehicle performance, optimize energy consumption, and improve user experiences.

Enhancing Travel Time Prediction for Intelligent Transportation Systems: A High-Resolution Origin–Destination-Based Approach with Multi-Dimensional Features

Accurate travel time prediction is essential for improving urban mobility, traffic management, and ride-hailing services. Traditional link- and path-based models face limitations due to data sparsity, segmentation errors, and computational inefficiencies. This study introduces an origin–destination (OD)-based travel time prediction framework leveraging high-resolution ride-hailing trajectory data. Unlike previous works, our approach systematically integrates spatiotemporal, quantified weather metrics and driver behavior clustering to enhance predictive accuracy. The proposed model employs a Back Propagation Neural Network (BPNN), which dynamically adjusts hyperparameters to improve generalization and mitigate overfitting. Empirical validation using ride-hailing data from Xi’an, China, demonstrates superior predictive performance, particularly for medium-range trips, achieving an RMSE of 202.89 s and a MAPE of 16.52%. Comprehensive ablation studies highlight the incremental benefits of incorporating spatiotemporal, weather, and behavioral features, showcasing their contributions to reducing prediction errors. While the model excels in moderate-speed scenarios, it exhibits limitations in short trips and low-speed cases due to data imbalance. Future research will enhance model robustness through data augmentation, real-time traffic integration, and scenario-specific adaptations. This study provides a scalable and adaptable travel time prediction framework, offering valuable insights for urban traffic management, dynamic route optimization, and sustainable mobility solutions within ITS.

Simulation and Experimental of PSMEID with Preview for Controlling Shock Vibration on UAV's Landing Gear

Several studies have been conducted on developing semi-active shock absorbers operating based on the principle of momentum exchange, more commonly known as Pre-Straining Spring Momentum Exchange Impact Damper (PSMEID). The initial concept of developing PSMEID was introduced as a customizable passive shock absorber system that can be transformed into an active damper by adding a simple mechatronic system to activate the vibration-damping process. This research initially simulated and conducted experimental studies on applying PSMEID to dampen shock vibrations on the landing gear of uncrewed aerial vehicles (UAVs) with the addition of active time prediction (preview). This study proves that PSMEID equipped with an active time prediction system reduces the amplitude of shock vibration acceleration during the landing process. Placing PSMEID on the landing gear side (suspended mass) effectively reduces the maximum vibration acceleration values. Through simulations, the variation of the fall height determines when it is appropriate to activate the PSMEID to reduce the amplitude of the vibration excursion. Activating the PSMEID just before the impact of the landing gear on the runway surface is dominant to isolate the vibration in the main mass. Meanwhile, experiments using a drop-test mechanism with varying drop heights show that PSMEID can dampen maximum acceleration even when activated a few moments before impact. The phenomena observed in the experimental results are consistent with simulation outcomes for specific prominent mass cases. However, this research emphasizes that PSMEID should be equipped with a sensor system with impact time prediction to achieve more optimal results, as demonstrated in this research experiment.

Prediction of Corrosion Initiation Time in Sustainable Concrete Blended with Microsilica and Portland Slag Cement Under Extreme Chloride Conditions

Prediction of Corrosion Initiation Time in Sustainable Concrete Blended with Microsilica and Portland Slag Cement Under Extreme Chloride Conditions

Deep learning methods for clinical workflow phase-based prediction of procedure duration: a benchmark study

This study evaluates the performance of deep learning models in the prediction of the end time of procedures performed in the cardiac catheterization laboratory (cath lab). We employed only the clinical phases derived from video analysis as input to the algorithms. Our results show that InceptionTime and LSTM-FCN yielded the most accurate predictions. InceptionTime achieves Mean Absolute Error (MAE) values below 5 min and Symmetric Mean Absolute Percentage Error (SMAPE) under 6% at 60-s sampling intervals. In contrast, LSTM with attention mechanism and standard LSTM models have higher error rates, indicating challenges in handling both long-term and short-term dependencies. CNN-based models, especially InceptionTime, excel at feature extraction across different scales, making them effective for time-series predictions. We also analyzed training and testing times. CNN models, despite higher computational costs, significantly reduce prediction errors. The Transformer model has the fastest inference time, making it ideal for real-time applications. An ensemble model derived by averaging the two best performing algorithms reported low MAE and SMAPE, although needing longer training. Future research should validate these findings across different procedural contexts and explore ways to optimize training times without losing accuracy. Integrating these models into clinical scheduling systems could improve efficiency in cath labs. Our research demonstrates that the models we implemented can form the basis of an automated tool, which predicts the optimal time to call the next patient with an average error of approximately 30 s. These findings show the effectiveness of deep learning models, especially CNN-based architectures, in accurately predicting procedure end times.

Research on the Prediction Problem of Outpatient Income Data Based on the Time Prediction Model

With the widespread application of internet big data, accurate Medical outpatient data prediction is crucial for improving the efficiency of medical management systems and optimizing resource allocation. This study focuses on the rehabilitation medicine ward of a certain hospital over the past five years, selecting a date, number of patients, total drug revenue, and daily ward income as basic input data. After basic data multiprocessing, an ARIMA time series model is established to predict and fill part of the outpatient income data for the rehabilitation medicine ward. At the same time, under the premise of ADF test stability, the expected data results obtained from the ARIMA model form a feature matrix. Furthermore, a Random Forest Classification Model is constructed to get the feature importance of the predicted data, and the feature importance is used as the weight of the ARIMA model. The weighted average method is used to calculate the predicted value of the overall outpatient income data for the rehabilitation medicine ward. The results show that the established random forest ensemble model prediction method improves the accuracy of medical outpatient prediction by 17.7%, compared to the traditional ARIMA model, it has higher prediction stability and accuracy, and can significantly improve the performance of data mining algorithms on medical datasets.