Mobility Prediction Research Articles

Mobile single-lead ECG atrial fibrillation prediction enhancement integrated by standard ECG algorithm with Deep learning model

Abstract Background Artificial intelligence (AI) using electrocardiogram (ECG) enabled to predict atrial fibrillation (AF) in patients without documented AF. Mobile single-lead ECG is more convenient to surveil cardiac rhythm with simple measurement. However, AI-enabled arrhythmia predictability by mobile ECG is limited due to single channel utilization and longer duration for arrhythmia diagnosis. We aimed to enhance mobile single-lead ECG AF prediction AI algorithm integrated with 12-lead ECG using deep learning model. Method Based on 552,372 12-lead ECG data of 318,321 patients, a statistical AF prediction model employing a deep-learning approach was constituted. Out of single-lead 13,509 ECGs from a total of 6,719 patients, we utilized a total of 10,287 normal sinus rhythm ECGs from 5,170 patients. Resnet structure was utilized to distinguish subtle changes of the vicinity of P-wave. Single-lead mobile ECGs were adjusted with noise filtering and segmented every 10 seconds. A random under-sampling was applied to reduce bias from data imbalance. Both 12-lead ECG and single-lead ECG were allocated to training, validation, testing datasets in a 6:2:2 ratio. Then, we conducted transfer learning using the standard 12-lead ECG’s deep learning model to improve performance of single-lead mobile ECG deep learning model. Results AF was annotated in 26,541 (4.8%) with 12-lead ECG whereas 1,443 (21.2%) with single-lead mobile ECG. The area under the curve (AUC) value for predicting AF was 0.910 with 12-lead ECG, and 0.742 with mobile ECG. The predictive performance of mobile ECG was 67.5% in accuracy, 64.2% in sensitivity and 66.8% in F1-score. The AUC value of mobile ECG after applying transfer learning based on 12-lead ECG for AF prediction was increased to 0.790 with accuracy of 73.8%, sensitivity of 65.5% and F1-score of 71.0%. Conclusion Integration with deep learning algorithm of standard 12-lead ECG significantly improved the model performance of single-lead ECG AF prediction model compared to single-lead mobile ECG only based model. Easy application of mobile ECG with enhanced AF predictability might serve a more convenient method as a pre-emptive assistive tool to provide probabilistic prediction for PAF screening rather than 12-lead ECG.

Proactive Scheduling for mmWave Wireless LANs

To cope with growing wireless bandwidth demand, millimeter wave (mmWave) communication has been identified as a promising technology to deliver Gbps throughput. However, due to the susceptibility of mmWave signals to blockage, applications can experience significant performance variability as users move around due to rapid and significant variation in channel conditions. In this context, proactive schedulers that make use of future data rate prediction have potential to bring a significant performance improvement as compared to traditional schedulers. In this work, we explore the possibility of proactive scheduling that uses mobility prediction and some knowledge of the environment to predict future channel conditions. We present both an optimal proactive scheduler, which is based on an integer linear programming formulation and provides an upper bound on proactive scheduling performance, and a greedy heuristic proactive scheduler that is suitable for practical implementation. Extensive simulation results show that proactive scheduling has the potential to increase average user data rate by up to 35% over the classic proportional fair scheduler without any loss of fairness and incurring only a small increase in jitter. The results also show that the efficient proactive heuristic scheduler achieves from 60% to 75% of the performance gains of the optimal proactive scheduler. Finally, the results show that proactive scheduling performance is sensitive to the quality of mobility prediction and, thus, use of state-of-the-art mobility prediction techniques will be necessary to realize its full potential.

Comparative Evaluation of Machine Learning Models for Mobile Phone Price Prediction: Assessing Accuracy, Robustness, and Generalization Performance

These days, mobile phones are the most commonly purchased goods. Thousands of new models with improved features, designs, and specifications are released yearly. An autonomous mobile price prediction system is required to assist customers in determining whether or not they can afford these devices. Many machine learning models exhibit varying performance degrees based on their architecture and learning properties. Ten widely used classifiers were assessed in this study: Logistic Regression (LR), Random Forest (RF), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Decision Tree (DT), Naïve Bayes (NB), Linear Discriminant Analysis (LDA), AdaBoost, and Light Gradient Boosting (LGB). The F1-score, recall, accuracy, and precision of these models were evaluated. According to the findings, the results indicated that LR, with its use of the Elastic Net parameter, outperformed the others with 96% accuracy, 97% precision, 94% recall, and 96% F1-score. Other models like XGBoost, LGB, and SVM also showed strong performance, whereas KNN had the poorest performance. The study highlights the importance of selecting the appropriate model for accurate mobile price prediction. Among all the machine learning used in this paper, the LR classifier outperforms the other state-of-the-art models because of the elastic Net parameter used for mobile phone price prediction.

Mobility Prediction via Rule-Enhanced Knowledge Graph

With the rapid development of location acquisition technologies, massive mobile trajectories have been collected and made available to us, which support a fantastic way of understanding and modeling individuals’ mobility. However, existing data-driven methods either fail to capture the long-range dependency or suffer from a high computational cost. To overcome these issues, we propose a knowledge-driven framework for mobility prediction, which leverages knowledge graphs (KG) to formulate the mobility prediction task into the KG completion problem through integrating the structured “knowledge” from the mobility data. However, most related mobility prediction works only focus on the structured information encoded in existing triples, which ignores the rich semantic information of relation paths composed of multiple relation triples. In this article, we apply a dedicated module to extract the supplementary semantic structure of paths in KG, which contributes to the interpretability and accuracy of our model. Specifically, the extracted rules are applied to capture the dependencies between relational facts. Moreover, by incorporating user information in the entity-relation space with the corresponding hyperplane, our method could capture diverse user mobility patterns and model the personal characteristics of users to improve the accuracy of mobility prediction. Extensive evaluations illustrate that our proposed model beats state-of-the-art mobility prediction algorithms, which verifies the superiority of utilizing logical rules and user hyperplanes. Our implementation code is available at https://github.com/tsinghua-fib-lab/RulekG-MobiPre.git

First-principles prediction of high carrier mobility for β-phase MX2N4 (M = Mo, W; X = Si, Ge) monolayers

The recently synthesized monolayer MoSi2N4 (Science 2020, 369, 367) exhibits exceptional environmental stability, a moderate band gap, and excellent mechanical properties, presenting exciting opportunities for the exploration of two-dimensional (2D) MX2Z4 materials. However, the low carrier mobility of α-phase MoSi2N4 significantly limits its potential applications in field-effect transistor (FET) devices. In this study, we systematically investigate the structural stability, elastic properties, and carrier mobility of a novel family of β-phase MX2N4 (M = Mo, W; X = Si, Ge) monolayers through first-principles calculations. Our findings reveal that these β-phase MX2N4 monolayers demonstrate remarkable dynamic, thermal, and mechanical stability. Specifically, we identify the MoSi2N4, MoGe2N4, WSi2N4, and WGe2N4 monolayers as semiconductors with band gaps of 2.70 eV, 1.57 eV, 3.12 eV, and 1.93 eV, respectively, as calculated using the HSE06 functional. Moreover, the MX2N4 monolayers exhibit significant elastic anisotropy, characterized by high ideal tensile strengths and a critical tensile strain exceeding 25%. Notably, the WGe2N4 monolayer displays exceptional anisotropic in-plane charge transport, achieving mobility levels of up to 104 cm2V− 1S− 1, surpassing those of the α-phase MX2N4 monolayers. These novel ternary monolayer structures have the potential to broaden the 2D MX2Z4 material family and emerge as promising candidates for applications in field-effect transistors.

Spectrum occupancy prediction using LSTM models for cognitive radio applications

ABSTRACT In recent days, mobile traffic prediction has become a prominent solution for spectrum management-related operations for the next-generation cellular networks in Cognitive Radio (CR) applications. To achieve this, the binary dataset has been created from the captured data by monitoring the spectrum activities of nine different Long Term Evolution (LTE) frequency channels. We propose a Long Short Term Memory (LSTM) based Spectrum Occupancy Prediction (SOP) approach for modelling infrastructure-based cellular traffic systems. The different types of LSTM models, such as Convolutional, Convolutional Neural Network (CNN), Stacked, and Bidirectional have been generated via offline training and tested for the created binary datasets. Moreover, the prediction performance evaluation of the generated LSTM models has been calculated using Mean Absolute Error (MAE). The pro- posed LSTM-based SOP model has achieved 2.5% higher prediction accuracy than the Auto-Regressive Integrated Moving Average (ARIMA) statistical model, accurately aligning the traffic trend with the actual samples.

Theoretical prediction of holes and electrons mobility in coronene and perchlorocoronene crystals: Application of the Marcus–Jortner–Levich Theory

The study focused on theoretically predicting charge carrier mobilities in coronene and perchlorocoronene crystals using the Marcus–Levich–Jortner approach. By analyzing charge transfer integrals and considering hopping pathways, significant anisotropy in charge transfer was observed. Density functional theory (DFT) calculations at the B3LYP-D3/6-311++G(d,p) and ωB97X-D/6-311++G(d,p) levels were employed to assess the impact of normal modes on reorganization energies and charge transfer rates. Results indicated that perchlorination reduced mobility for both holes and electrons. Notably, maximum mobilities were observed at the melting points: 0.0120 cm2 V−1 s−1 for electrons and 0.0798 cm2 V−1 s−1 for holes in coronene at 455 K, and 0.0032 cm2 V−1 s−1 for electrons and 0.0426 cm2 V−1 s−1 for holes in perchlorocoronene at 560 K. This suggests that both crystals exhibit favorable charge transport characteristics near their respective melting points.

Impact of heteroatoms and chemical functionalisation on crystal structure and carrier mobility of organic semiconductors

The substitution of heteroatoms and the functionalisation of molecules are established strategies in chemical synthesis. They target the precise tuning of the electronic properties of hydrocarbon molecules to improve their performance in various applications and increase their versatility. Modifications to the molecular structure often lead to simultaneous changes in the morphology such as different crystal structures. These changes can have a stronger and unpredictable impact on the targeted property. The complex relationships between substitution/functionalization in chemical synthesis and the resulting modifications of properties in thin films or crystals are difficult to predict and remain elusive. Here we address these effects for charge carrier transport in organic crystals by combining simulations of carrier mobilities with crystal structure prediction based on density functional theory and density functional tight binding theory. This enables the prediction of carrier mobilities based solely on the molecular structure and allows for the investigation of chemical modifications prior to synthesis and characterisation. Studying nine specific molecules with tetracene and rubrene as reference compounds along with their combined modifications of the molecular cores and additional functionalisations, we unveil systematic trends for the carrier mobilities of their polymorphs. The positive effect of phenyl groups that is responsible for the marked differences between tetracene and rubrene can be transferred to other small molecules such as NDT and NBT leading to a mobility increase by large factors of about five.

Dynamic predictability and activity-location contexts in human mobility.

Human travelling behaviours are markedly regular, to a large extent predictable, and mostly driven by biological necessities and social constructs. Not surprisingly, such predictability is influenced by an array of factors ranging in scale from individual preferences and choices, through social groups and households, all the way to the global scale, such as mobility restrictions in response to external shocks such as pandemics. In this work, we explore how temporal, activity and location variations in individual-level mobility-referred to as predictability states-carry a large degree of information regarding the nature of mobility regularities at the population level. Our findings indicate the existence of contextual and activity signatures in predictability states, suggesting the potential for a more nuanced approach to estimating both short-term and higher-order mobility predictions. The existence of location contexts, in particular, serves as a parsimonious estimator for predictability patterns even in the case of low resolution and missing data.

Combining the spatiotemporal mobility patterns and MMC for next location prediction of fake base stations

The spatiotemporal mobility patterns and next location prediction of fake base stations (FBS) provide important technical support for the police to prevent spam messages from FBS. However, due to the difficulty in locating their real-time locations, our understanding of the mobility patterns and predictability of FBS is still limited. Based on the crowdsourced spam data, we extract the time and potential locations of FBS and propose a Tucker-MMC method that combines Tucker decomposition with a Mobility Markov Chain (MMC) model to investigate the mobility patterns and predictability of FBS sending spam messages. First, we utilize Tucker decomposition to reflect the spatial and temporal preferences during the movement of the corresponding FBS. Then the mobility regularity and the theoretical maximum predictability of the FBS trajectories with similar mobility preferences are analyzed by entropy and Fano's inequality. A Tucker-MMC is also established for the next location prediction. The results using the spam dataset in Beijing show that the accuracy of Tucker-MMC is more than double that of the MMC. The accuracy of the actual location prediction model is more likely to approach the theoretical maximum predictability when FBS send spam messages in a shorter time, shorter transfer distance, and smaller access range.

DeepAGS: Deep learning with activity, geography and sequential information in predicting an individual's next trip destination

AbstractIndividual mobility is driven by activities and thus restricted geographically, especially for trip destination prediction in public transport. Existing statistical learning based models focus on extracting mobility regularity in predicting an individual's mobility. However, they are limited in modeling varied spatial mobility patterns driven by the same activity (e.g. an individual may travel to different locations for shopping). The paper proposes a deep learning model with activity, geographic and sequential (DeepAGS) information in predicting an individual's next trip destination in public transport. DeepAGS models the semantic features of activity and geography by using word embedding and graph convolutional network. An adaptive neural fusion gate mechanism is proposed to dynamically fuse the mobility activity and geographical information given the current trip information. Besides, DeepAGS uses the gated recurrent unit to capture the temporal mobility regularity. The approach is validated by using a real‐world smartcard dataset in urban railway systems and comparing with state‐of‐the‐art models. The results show that the proposed model outperforms its peers in terms of accuracy and robustness by effectively integrating the activity and geographical information relevant to a trip context. Also, we illustrate and verify the working mechanism of the DeepAGS model using the synthetic data constructed using real‐world data. The DeepAGS model captures both the activity and geographic information of hidden mobility activities and thus could be potentially applicable to other mobility prediction tasks, such as bus trip destinations and individual GPS locations.

A Hybrid Localization Algorithm for Underwater Nodes Based on Neural Network and Mobility Prediction

A Hybrid Localization Algorithm for Underwater Nodes Based on Neural Network and Mobility Prediction

Accelerating the calculation of electron-phonon coupling strength with machine learning.

The calculation of electron-phonon couplings (EPCs) is essential for understanding various fundamental physical properties, including electrical transport, optical and superconducting behaviors in materials. However, obtaining EPCs through fully first-principles methods is notably challenging, particularly for large systems or when employing advanced functionals. Here we introduce a machine learning framework to accelerate EPC calculations by utilizing atomic orbital-based Hamiltonian matrices and gradients predicted by an equivariant graph neural network. We demonstrate that our method not only yields EPC values in close agreement with first-principles results but also enhances calculation efficiency by several orders of magnitude. Application to GaAs using the Heyd-Scuseria-Ernzerhof functional reveals the necessity of advanced functionals for accurate carrier mobility predictions, while for the large Kagome crystal CsV3Sb5, our framework reproduces the experimentally observed double domes in pressure-induced superconducting phase diagrams. This machine learning framework offers a powerful and efficient tool for the investigation of diverse EPC-related phenomena in complex materials.

Predictability of human mobility during the COVID-19 pandemic in the United States.

Human mobility is fundamental to a range of applications including epidemic control, urban planning, and traffic engineering. While laws governing individual movement trajectories and population flows across locations have been extensively studied, the predictability of population-level mobility during the COVID-19 pandemic driven by specific activities such as work, shopping, and recreation remains elusive. Here we analyze mobility data for six place categories at the US county level from 2020 February 15 to 2021 November 23 and measure how the predictability of these mobility metrics changed during the COVID-19 pandemic. We quantify the time-varying predictability in each place category using an information-theoretic metric, permutation entropy. We find disparate predictability patterns across place categories over the course of the pandemic, suggesting differential behavioral changes in human activities perturbed by disease outbreaks. Notably, predictability change in foot traffic to residential locations is mostly in the opposite direction to other mobility categories. Specifically, visits to residences had the highest predictability during stay-at-home orders in March 2020, while visits to other location types had low predictability during this period. This pattern flipped after the lifting of restrictions during summer 2020. We identify four key factors, including weather conditions, population size, COVID-19 case growth, and government policies, and estimate their nonlinear effects on mobility predictability. Our findings provide insights on how people change their behaviors during public health emergencies and may inform improved interventions in future epidemics.

Next-cell and mobility prediction in new generation cellular systems based on convolutional neural networks and encoding mobility data as images

Mobility prediction has been a popular research topic for many decades. With the advent of new generation technologies (5G and beyond) and smaller coverage cells, hand-over operations have become more frequent. Cellular system companies are therefore taking increasing interest in using the available predictive information on node movements to optimize and manage their bandwidth resources. In particular, the main challenging scope of our contribution consists in solving the issue of reliable next-cell prediction, aimed to call dropping probability minimization. In addition, our proposal is based on the innovative concept of mobility data to image encoding. The scheme is able to a-priori determine the next visited cells during host movements by applying a convolutional neural approach to mobility images. The power of machine learning is used to advantage, and highly accurate image classification is achieved for mobility prediction. We performed numerous simulation campaigns related to next-cell prediction in mobile cellular environments, obtaining very satisfactory results by the application of convolutional neural networks, which have an impressive history of effectiveness with image classification problems. The trained network has been associated to each coverage cell and the prediction accuracy has been evaluated.

Cross- and Context-Aware Attention Based Spatial-Temporal Graph Convolutional Networks for Human Mobility Prediction

The COVID-19 pandemic has dramatically transformed human mobility patterns. Therefore, human mobility prediction for the “new normal” is crucial to infrastructure redesign, emergency management, and urban planning post the pandemic. This paper aims to predict people’s number of visits to various locations in New York City using COVID and mobility data in the past two years. To quantitatively model the impact of COVID cases on human mobility patterns and predict mobility patterns across the pandemic period, this paper develops a model CCAAT-GCN ( C ross- and C ontext- A ttention based Spatial-Temporal G raph C onvolutional N etworks). The proposed model is validated using SafeGraph data in New York City from August 2020 to April 2022. A rich set of baselines are performed to demonstrate the performance of our proposed model. Results demonstrate the superior performance of our proposed method. Also, the attention matrix learned by our model exhibits a strong alignment with the COVID-19 situation and the points of interest within the geographic region. This alignment suggests that the model effectively captures the intricate relationships between COVID-19 case rates and human mobility patterns. The developed model and findings can offer insights into the mobility pattern prediction for future disruptive events and pandemics, so as to assist with emergency preparedness for planners, decision-makers and policymakers.

Dynamic Spectrum Optimization for Internet-of-Vehicles with Deep-Learning-Based Mobility Prediction

Dynamic Spectrum Optimization for Internet-of-Vehicles with Deep-Learning-Based Mobility Prediction

SecDM: A Secure and Lossless Human Mobility Prediction System

SecDM: A Secure and Lossless Human Mobility Prediction System

Predictability in Human Mobility: From Individual to Collective (Vision Paper)

Human mobility is the foundation of urban dynamics and its prediction significantly benefits various downstream location-based services. Nowadays, while deep learning approaches are dominating the mobility prediction field where various model architectures/designs are continuously updating to push up the prediction accuracy, there naturally arises a question: whether these models are sufficiently good to reach the best possible prediction accuracy? To answer this question, predictability study is a method that quantifies the inherent regularities of the human mobility data and links the result to that limit. Mainstream predictability studies achieve this by analyzing the individual trajectories and merging all individual results to obtain an upper bound. However, the multiple individuals composing the city are not totally independent and the individual behavior is heavily influenced by its implicit or explicit surroundings. Therefore, the collective factor should be considered in the mobility predictability measurement, which has not been addressed before. This vision paper points out this concern and envisions a few potential research problems along such an individual-to-collective transition from both data and methodology aspects. We hope the discussion in this paper sheds some light on the human mobility predictability community.

Mobile traffic prediction with attention-based hybrid deep learning

Mobile data consumption is ever-increasing due to the continuous emergence of bandwidth-hungry mobile applications as well as the drastic increase in number of mobile terminals resulted from the pervasive adoption of Internet of Things. In the 5G+ era, network operators have to deal with new technical challenges on network management, resource optimization and mobility scheduling, which require the ability to perform efficient mobile traffic prediction. Traditional prediction methods using convex optimization, Markov model and linear equations cannot describe the various nonlinearity changes of network traffic. In this paper, we propose a novel traffic prediction algorithm for mobile networks by introducing a hybrid deep learning method equipped with the attention mechanism, which analyzes the traffic in cellular zones to determine the spatiotemporal characteristics in the residential community. The hybrid deep learning model includes autoregressive integrated moving average mode (ARIMA) preprocessing with decomposition ability for time series, external spatial feature variables, sequence-to-sequence (seq2seq; attention-based CNN-LSTM) pretraining and extreme gradient boosting (XGBoost) optimization processing for integrating nonlinear relationships. The CNN method shows good nonlinear generalization ability to mine deep features of community traffic and capture the nonlinear trend. Our proposed approach first decomposes the community traffic data into linear and residual parts through ARIMA and then seizes the deep features of community traffic through the constructed pretraining seq2seq framework. Finally, the pretraining outcomes are further adjusted by the XGBoost method. Simulation results show that the proposed solution improves the accuracy of prediction apparently.