Spatio-temporal prediction of the COVID-19 pandemic in US counties: modeling with a deep LSTM neural network

Enhanced SARS-CoV-2 case prediction using public health data and machine learning models

Objectives

The goal of this study is to propose and test a scalable framework for machine learning (ML) algorithms to predict near-term severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) cases by incorporating and evaluating the impact of real-time dynamic public health data.

Materials and Methods

Data used in this study include patient-level results, procurement, and location information of all SARS-CoV-2 tests reported in West Virginia as part of their mandatory reporting system from January 2021 to March 2022. We propose a method for incorporating and comparing widely available public health metrics inside of a ML framework, specifically a long-short-term memory network, to forecast SARS-CoV-2 cases across various feature sets.

Results

Our approach provides better prediction of localized case counts and indicates the impact of the dynamic elements of the pandemic on predictions, such as the influence of the mixture of viral variants in the population and variable testing and vaccination rates during various eras of the pandemic.

Discussion

Utilizing real-time public health metrics, including estimated Rt from multiple SARS-CoV-2 variants, vaccination rates, and testing information, provided a significant increase in the accuracy of the model during the Omicron and Delta period, thus providing more precise forecasting of daily case counts at the county level. This work provides insights on the influence of various features on predictive performance in rural and non-rural areas.

Conclusion

Our proposed framework incorporates available public health metrics with operational data on the impact of testing, vaccination, and current viral variant mixtures in the population to provide a foundation for combining dynamic public health metrics and ML models to deliver forecasting and insights in healthcare domains. It also shows the importance of developing and deploying ML frameworks in rural settings.

A brief review and scientometric analysis on ensemble learning methods for handling COVID-19

Numerous efforts and research have been conducted worldwide to combat the coronavirus disease 2019 (COVID-19) pandemic. In this regard, some researchers have focused on deep and machine-learning approaches to discover more about this disease. There have been many articles on using ensemble learning methods for COVID-19 detection. Still, there seems to be no scientometric analysis or a brief review of these researches. Hence, a combined method of scientometric analysis and brief review was used to study the published articles that employed an ensemble learning approach to detect COVID-19. This research used both methods to overcome their limitations, leading to enhanced and reliable outcomes. The related articles were retrieved from the Scopus database. Then a two-step procedure was employed. A concise review of the collected articles was conducted. Then they underwent scientometric and bibliometric analyses. The findings revealed that convolutional neural network (CNN) is the mostly employed algorithm, while support vector machine (SVM), random forest, Resnet, DenseNet, and visual geometry group (VGG) were also frequently used. Additionally, China has had a significant presence in the numerous top-ranking categories of this field of research. Both study phases yielded valuable results and rankings.

A data-driven epidemic model with human mobility and vaccination protection for COVID-19 prediction

Epidemiological models allow for quantifying the dynamic characteristics of large-scale outbreaks. However, capturing detailed and accurate epidemiological information often requires consideration of multiple kinetic mechanisms and parameters. Due to the uncertainty of pandemic evolution, such as pathogen variation, host immune response and changes in mitigation strategies, the parameter evaluation and state prediction of complex epidemiological models are challenging. Here, we develop a data-driven epidemic model with a generalized SEIR mechanistic structure that includes new compartments, human mobility and vaccination protection. To address the issue of model complexity, we embed the epidemiological model dynamics into physics-informed neural networks (PINN), taking the observed series of time instances as direct input of the network to simultaneously infer unknown parameters and unobserved dynamics of the underlying model. Using actual data during the COVID-19 outbreak in Australia, Israel, and Switzerland, our model framework demonstrates satisfactory performance in multi-step ahead predictions compared to several benchmark models. Moreover, our model infers time-varying parameters such as transmission rates, hospitalization ratios, and effective reproduction numbers, as well as calculates the latent period and asymptomatic infection count, which are typically unreported in public data. Finally, we employ the proposed data-driven model to analyze the impact of different mitigation strategies on COVID-19.

Application of optimized LSTM in prediction of the cumulative confirmed cases of COVID-19

This paper proposes an optimized Long Short-Term Memory (LSTM+) model for predicting cumulative confirmed cases of COVID-19 in Germany, the UK, Italy, and Japan. The LSTM+ model incorporates two key optimizations: (1) fine-adjustment of parameters and (2) a 're-prediction' process that utilizes the latest prediction results from the previous iteration. The performance of the LSTM+ model is evaluated and compared with that of Backpropagation (BP) and traditional LSTM models. The results demonstrate that the LSTM+ model significantly outperforms both BP and LSTM models, achieving a Mean Absolute Percentage Error (MAPE) of less than 0.6%. Additionally, two illustrative examples employing the LSTM+ model further validate its general applicability and practical performance for predicting cumulative confirmed COVID-19 cases.

Transmission dynamics informed neural network with application to COVID-19 infections

Since the end of 2019 the COVID-19 repeatedly surges with most countries/territories experiencing multiple waves, and mechanism-based epidemic models played important roles in understanding the transmission mechanism of multiple epidemic waves. However, capturing temporal changes of the transmissibility of COVID-19 during the multiple waves keeps ill-posed problem for traditional mechanism-based epidemic compartment models, because that the transmission rate is usually assumed to be specific piecewise functions and more parameters are added to the model once multiple epidemic waves involved, which poses a huge challenge to parameter estimation. Meanwhile, data-driven deep neural networks fail to discover the driving factors of repeated outbreaks and lack interpretability. In this study, aiming at developing a data-driven method to project time-dependent parameters but also merging the advantage of mechanism-based models, we propose a transmission dynamics informed neural network (TDINN) by encoding the SEIRD compartment model into deep neural networks. We show that the proposed TDINN algorithm performs very well when fitting the COVID-19 epidemic data with multiple waves, where the epidemics in the United States, Italy, South Africa, and Kenya, and several outbreaks the Omicron variant in China are taken as examples. In addition, the numerical simulation shows that the trained TDINN can also perform as a predictive model to capture the future development of COVID-19 epidemic. We find that the transmission rate inferred by the TDINN frequently fluctuates, and a feedback loop between the epidemic shifting and the changes of transmissibility drives the occurrence of multiple waves. We observe a long response delay to the implementation of control interventions in the four countries, while the decline of the transmission rate in the outbreaks in China usually happens once the implementation of control interventions. The further simulation show that 17 days' delay of the response to the implementation of control interventions lead to a roughly four-fold increase in daily reported cases in one epidemic wave in Italy, which suggest that a rapid response to policies that strengthen control interventions can be effective in flattening the epidemic curve or avoiding subsequent epidemic waves. We observe that the transmission rate in the outbreaks in China is already decreasing before enhancing control interventions, providing the evidence that the increasing of the epidemics can drive self-conscious behavioural changes to protect against infections.

Pandemic vulnerability index of US cities: A hybrid knowledge-based and data-driven approach

Cities become mission-critical zones during pandemics and it is vital to develop a better understanding of the factors that are associated with infection levels. The COVID-19 pandemic has impacted many cities severely; however, there is significant variance in its impact across cities. Pandemic infection levels are associated with inherent features of cities (e.g., population size, density, mobility patterns, socioeconomic condition, and health & environment), which need to be better understood. Intuitively, the infection levels are expected to be higher in big urban agglomerations, but the measurable influence of a specific urban feature is unclear. The present study examines 41 variables and their potential influence on the incidence of COVID-19 infection cases. The study uses a multi-method approach to study the influence of variables, classified as demographic, socioeconomic, mobility and connectivity, urban form and density, and health and environment dimensions. This study develops an index dubbed the pandemic vulnerability index at city level (PVI-CI) for classifying the pandemic vulnerability levels of cities, grouping them into five vulnerability classes, from very high to very low. Furthermore, clustering and outlier analysis provides insights on the spatial clustering of cities with high and low vulnerability scores. This study provides strategic insights into levels of influence of key variables upon the spread of infections, along with an objective ranking for the vulnerability of cities. Thus, it provides critical wisdom needed for urban healthcare policy and resource management. The calculation method for the pandemic vulnerability index and the associated analytical process present a blueprint for the development of similar indices for cities in other countries, leading to a better understanding and improved pandemic management for urban areas, and more resilient planning for future pandemics in cities across the world.

Evidence-driven spatiotemporal COVID-19 hospitalization prediction with Ising dynamics

In this work, we aim to accurately predict the number of hospitalizations during the COVID-19 pandemic by developing a spatiotemporal prediction model. We propose HOIST, an Ising dynamics-based deep learning model for spatiotemporal COVID-19 hospitalization prediction. By drawing the analogy between locations and lattice sites in statistical mechanics, we use the Ising dynamics to guide the model to extract and utilize spatial relationships across locations and model the complex influence of granular information from real-world clinical evidence. By leveraging rich linked databases, including insurance claims, census information, and hospital resource usage data across the U.S., we evaluate the HOIST model on the large-scale spatiotemporal COVID-19 hospitalization prediction task for 2299 counties in the U.S. In the 4-week hospitalization prediction task, HOIST achieves 368.7 mean absolute error, 0.6 [Formula: see text] and 0.89 concordance correlation coefficient score on average. Our detailed number needed to treat (NNT) and cost analysis suggest that future COVID-19 vaccination efforts may be most impactful in rural areas. This model may serve as a resource for future county and state-level vaccination efforts.

Transfer-recursive-ensemble learning for multi-day COVID-19 prediction in India using recurrent neural networks

The COVID-19 pandemic has put a huge challenge on the Indian health infrastructure. With a larger number of people getting affected during the second wave, hospitals were overburdened, running out of supplies and oxygen. Hence, predicting new COVID-19 cases, new deaths, and total active cases multiple days in advance can aid better utilization of scarce medical resources and prudent pandemic-related decision-making. The proposed method uses gated recurrent unit networks as the main predicting model. A study is conducted by building four models pre-trained on COVID-19 data from four different countries (United States of America, Brazil, Spain, and Bangladesh) and fine-tuned on India's data. Since the four countries chosen have experienced different types of infection curves, the pre-training provides a transfer learning to the models incorporating diverse situations into account. Each of the four models then gives 7-day ahead predictions using the recursive learning method for the Indian test data. The final prediction comes from an ensemble of the predictions of the different models. This method with two countries, Spain and Bangladesh, is seen to achieve the best performance amongst all the combinations as well as compared to other traditional regression models.

Incorporating variant frequencies data into short-term forecasting for COVID-19 cases and deaths in the USA: a deep learning approach

BACKGROUND

Since the US reported its first COVID-19 case on January 21, 2020, the science community has been applying various techniques to forecast incident cases and deaths. To date, providing an accurate and robust forecast at a high spatial resolution has proved challenging, even in the short term.

METHOD

Here we present a novel multi-stage deep learning model to forecast the number of COVID-19 cases and deaths for each US state at a weekly level for a forecast horizon of 1-4 weeks. The model is heavily data driven, and relies on epidemiological, mobility, survey, climate, demographic, and SARS-CoV-2 variant frequencies data. We implement a rigorous and robust evaluation of our model-specifically we report on weekly performance over a one-year period based on multiple error metrics, and explicitly assess how our model performance varies over space, chronological time, and different outbreak phases.

FINDINGS

The proposed model is shown to consistently outperform the CDC ensemble model for all evaluation metrics in multiple spatiotemporal settings, especially for the longer-term (3 and 4 weeks ahead) forecast horizon. Our case study also highlights the potential value of variant frequencies data for use in short-term forecasting to identify forthcoming surges driven by new variants.

INTERPRETATION

Based on our findings, the proposed forecasting framework improves upon the available state-of-the-art forecasting tools currently used to support public health decision making with respect to COVID-19 risk.

FUNDING

This work was funded the NSF Rapid Response Research (RAPID) grant Award ID 2108526 and the CDC Contract #75D30120C09570.

Sensor Data Prediction in Missile Flight Tests

Sensor data from missile flights are highly valuable, as a test requires considerable resources, but some sensors may be detached or fail to collect data. Remotely acquired missile sensor data are incomplete, and the correlations between the missile data are complex, which results in the prediction of sensor data being difficult. This article proposes a deep learning-based prediction network combined with the wavelet analysis method. The proposed network includes an imputer network and a prediction network. In the imputer network, the data are decomposed using wavelet transform, and the generative adversarial networks assist the decomposed data in reproducing the detailed information. The prediction network consists of long short-term memory with an attention and dilation network for accurate prediction. In the test, the actual sensor data from missile flights were used. For the performance evaluation, the test was conducted from the data with no missing values to the data with five different missing rates. The test results showed that the proposed system predicts the missile sensor most accurately in all cases. In the frequency analysis, the proposed system has similar frequency responses to the actual sensors and showed that the proposed system accurately predicted the sensors in both tendency and frequency aspects.

Simulating and Forecasting the COVID-19 Spread in a U.S. Metropolitan Region with a Spatial SEIR Model

The global COVID-19 pandemic has taken a heavy toll on health, social, and economic costs since the end of 2019. Predicting the spread of a pandemic is essential to developing effective intervention policies. Since the beginning of this pandemic, many models have been developed to predict its pathways. However, the majority of these models assume homogeneous dynamics over the geographic space, while the pandemic exhibits substantial spatial heterogeneity. In addition, spatial interaction among territorial entities and variations in their magnitude impact the pandemic dynamics. In this study, we used a spatial extension of the SEIR-type epidemiological model to simulate and predict the 4-week number of COVID-19 cases in the Charlotte-Concord-Gastonia Metropolitan Statistical Area (MSA), USA. We incorporated a variety of covariates, including mobility, pharmaceutical, and non-pharmaceutical interventions, demographics, and weather data to improve the model's predictive performance. We predicted the number of COVID-19 cases for up to four weeks in the 10 counties of the studied MSA simultaneously over the time period 29 March 2020 to 13 March 2021, and compared the results with the reported number of cases using the root-mean-squared error (RMSE) metric. Our results highlight the importance of spatial heterogeneity and spatial interactions among locations in COVID-19 pandemic modeling.

Spatio-temporal variation of Covid-19 health outcomes in India using deep learning based models

Deep learning methods have become the state of the art for spatio-temporal predictive analysis in a wide range of fields, including environmental management, public health, urban planning, pollution monitoring, and so on. Despite the fact that a variety of powerful deep learning-based models can address various problem-specific issues in different research domain, it has been found that no single optimal model can outperform everywhere. Now, in the last two years, various deep learning-based studies have provided a variety of best-performing techniques for predicting COVID-19 health outcomes. In this context, this study attempts to perform a case study that investigates the spatio-temporal variation in the performance of deep-learning-based methods for predicting COVID-19 health outcomes in India. Various widely applied deep learning models namely CNN (convolutional neural network), RNN (recurrent neural network), Vanilla LSTM (long short-term memory), LSTM Autoencoder, and Bidirectional LSTM are considered to investigate their spatio-temporal performance variation. The effectiveness of the models is assessed using various metrics based on COVID-19 mortality time-series from 36 states and union territories of India.

A systematic review of COVID-19 transport policies and mitigation strategies around the globe

This paper reports a Scopus-based systematic literature review of a wide variety of transportation policies and mitigation strategies that have been conducted around the world to minimize COVID-19 contagion risk in transportation systems. The review offers a representative coverage of countries across all continents of the planet, as well as among representative climate regions - as weather is an important factor to consider. The readership interested in policies and mitigation strategies is expected to involve a wide range of actors, each involving a particular application context; hence, the literature is also characterized by key attributes such as: transportation mode; actor (users, operators, government, industry); jurisdiction (national, provincial, city, neighborhood); and area of application (planning, regulation, operations, research, incentives). An in-depth analysis of the surveyed literature is then reported, focusing first on condensing the literature into 151 distinct policies and strategies, which are subsequently categorized into 25 broad categories that are discussed at length. The compendium and discussion of strategies and policies reported not only provide comprehensive guidelines to inform various courses of action for decision-makers, planners, and social communicators, but also emphasize on future work and the potential of some of these strategies to be the precursors of meaningful, more sustainable behavioral changes in future mobility patterns.

COVID-19 forecasting using new viral variants and vaccination effectiveness models

Recently, a high number of daily positive COVID-19 cases have been reported in regions with relatively high vaccination rates; hence, booster vaccination has become necessary. In addition, infections caused by the different variants and correlated factors have not been discussed in depth. With large variabilities and different co-factors, it is difficult to use conventional mathematical models to forecast the incidence of COVID-19. Machine learning based on long short-term memory was applied to forecasting the time series of new daily positive cases (DPC), serious cases, hospitalized cases, and deaths. Data acquired from regions with high rates of vaccination, such as Israel, were blended with the current data of other regions in Japan such that the effect of vaccination was considered in efficient manner. The protection provided by symptomatic infection was also considered in terms of the population effectiveness of vaccination as well as the vaccination protection waning effect and ratio and infectivity of different viral variants. To represent changes in public behavior, public mobility and interactions through social media were also included in the analysis. Comparing the observed and estimated new DPC in Tel Aviv, Israel, the parameters characterizing vaccination effectiveness and the waning protection from infection were well estimated; the vaccination effectiveness of the second dose after 5 months and the third dose after two weeks from infection by the delta variant were 0.24 and 0.95, respectively. Using the extracted parameters regarding vaccination effectiveness, DPC in three major prefectures of Japan were replicated. The key factor influencing the prevention of COVID-19 transmission is the vaccination effectiveness at the population level, which considers the waning protection from vaccination rather than the percentage of fully vaccinated people. The threshold of the efficiency at the population level was estimated as 0.3 in Tel Aviv and 0.4 in Tokyo, Osaka, and Aichi. Moreover, a weighting scheme associated with infectivity results in more accurate forecasting by the infectivity model of viral variants. Results indicate that vaccination effectiveness and infectivity of viral variants are important factors in future forecasting of DPC. Moreover, this study demonstrate a feasible way to project the effect of vaccination using data obtained from other country.

Associations between COVID-19 Pandemic, Lockdown Measures and Human Mobility: Longitudinal Evidence from 86 Countries

Recognizing an urgent need to understand the dynamics of the pandemic’s severity, this longitudinal study is conducted to explore the evolution of complex relationships between the COVID-19 pandemic, lockdown measures, and social distancing patterns in a diverse set of 86 countries. Collecting data from multiple sources, a structural equation modeling (SEM) technique is applied to understand the interdependencies between independent variables, mediators, and dependent variables. Results show that lockdown and confinement measures are very effective to reduce human mobility at retail and recreation facilities, transit stations, and workplaces and encourage people to stay home and thereby control COVID-19 transmission at critical times. The study also found that national contexts rooted in socioeconomic and institutional factors influence social distancing patterns and severity of the pandemic, particularly with regard to the vulnerability of people, treatment costs, level of globalization, employment distribution, and degree of independence in society. Additionally, this study portrayed a mutual relationship between the COVID-19 pandemic and human mobility. A higher number of COVID-19 confirmed cases and deaths reduces human mobility and the countries with reduced personal mobility have experienced a deepening of the severity of the pandemic. However, the effect of mobility on pandemic severity is stronger than the effect of pandemic situations on mobility. Overall, the study displays considerable temporal changes in the relationships between independent variables, mediators, and dependent variables considering pandemic situations and lockdown regimes, which provides a critical knowledge base for future handling of pandemics. It has also accommodated some policy guidelines for the authority to control the transmission of COVID-19.

A Review on Deep Sequential Models for Forecasting Time Series Data

Deep sequential (DS) models are extensively employed for forecasting time series data since the dawn of the deep learning era, and they provide forecasts for the values required in subsequent time steps. DS models, unlike other traditional statistical models for forecasting time series data, can learn hidden patterns in temporal sequences and have the memorizing data from prior time points. Given the widespread usage of deep sequential models in several domains, a comprehensive study describing their applications is necessary. This work presents a comprehensive review of contemporary deep learning time series models, their performance in diverse domains, and an investigation of the models that were employed in various applications. Three deep sequential models, namely, artificial neural network (ANN), long short-term memory (LSTM), and temporal-conventional neural network (TCNN) along with their applications for forecasting time series data, are elaborated. We showed a comprehensive comparison between such models in terms of application fields, model structure and activation functions, optimizers, and implementation, with a goal of learning more about the optimal model used. Furthermore, the challenges and perspectives of future development of deep sequential models are presented and discussed. We conclude that the LSTM model is widely employed, particularly in the form of a hybrid model, in which the most accurate predictions are made when the shape of hybrids is used as the model.

FedSGDCOVID: Federated SGD COVID-19 Detection under Local Differential Privacy Using Chest X-ray Images and Symptom Information

Coronavirus (COVID-19) has created an unprecedented global crisis because of its detrimental effect on the global economy and health. COVID-19 cases have been rapidly increasing, with no sign of stopping. As a result, test kits and accurate detection models are in short supply. Early identification of COVID-19 patients will help decrease the infection rate. Thus, developing an automatic algorithm that enables the early detection of COVID-19 is essential. Moreover, patient data are sensitive, and they must be protected to prevent malicious attackers from revealing information through model updates and reconstruction. In this study, we presented a higher privacy-preserving federated learning system for COVID-19 detection without sharing data among data owners. First, we constructed a federated learning system using chest X-ray images and symptom information. The purpose is to develop a decentralized model across multiple hospitals without sharing data. We found that adding the spatial pyramid pooling to a 2D convolutional neural network improves the accuracy of chest X-ray images. Second, we explored that the accuracy of federated learning for COVID-19 identification reduces significantly for non-independent and identically distributed (Non-IID) data. We then proposed a strategy to improve the model's accuracy on Non-IID data by increasing the total number of clients, parallelism (client-fraction), and computation per client. Finally, for our federated learning model, we applied a differential privacy stochastic gradient descent (DP-SGD) to improve the privacy of patient data. We also proposed a strategy to maintain the robustness of federated learning to ensure the security and accuracy of the model.

COVIDNearTerm: A simple method to forecast COVID-19 hospitalizations

Introduction

COVID-19 has caused tremendous death and suffering since it first emerged in 2019. Soon after its emergence, models were developed to help predict the course of various disease metrics, and these models have been relied upon to help guide public health policy.

Methods

Here we present a method called COVIDNearTerm to "forecast" hospitalizations in the short term, two to four weeks from the time of prediction. COVIDNearTerm is based on an autoregressive model and utilizes a parametric bootstrap approach to make predictions. It is easy to use as it requires only previous hospitalization data, and there is an open-source R package that implements the algorithm. We evaluated COVIDNearTerm on San Francisco Bay Area hospitalizations and compared it to models from the California COVID Assessment Tool (CalCAT).

Results

We found that COVIDNearTerm predictions were more accurate than the CalCAT ensemble predictions for all comparisons and any CalCAT component for a majority of comparisons. For instance, at the county level our 14-day hospitalization median absolute percentage errors ranged from 16 to 36%. For those same comparisons, the CalCAT ensemble errors were between 30 and 59%.

Conclusion

COVIDNearTerm is a simple and useful tool for predicting near-term COVID-19 hospitalizations.

Post-pandemic COVID-19 estimated and forecasted hotspots in the Association of Southeast Asian Nations (ASEAN) countries in connection to vaccination rate

After a two-year pandemic, coronavirus disease 2019 (COVID-19) is still a serious public health problem and economic stability worldwide, particularly in the Association of Southeast Asian Nations (ASEAN) countries. The objective of this study was to identify the wave periods, provide an accurate space-time forecast of COVID-19 disease and its relationship to vaccination rates. We combined a hierarchical Bayesian pure spatiotemporal model and locally weighted scatterplot smoothing techniques to identify the wave periods and to provide weekly COVID-19 forecasts for the period 15 December 2021 to 5 January 2022 and to identify the relationship between the COVID-19 risk and the vaccination rate. We discovered that each ASIAN country had a unique COVID-19 time wave and duration. Additionally, we discovered that the number of COVID-19 cases was quite low and that no weekly hotspots were identified during the study period. The vaccination rate showed a nonlinear relationship with the COVID-19 risk, with a different temporal pattern for each ASEAN country. We reached the conclusion that vaccination, in comparison to other interventions, has a large influence over a longer time span.

A multi-sensory stimulating attention model for cities' taxi service demand prediction

Taxi demand forecasting is crucial to building an efficient transportation system in a smart city. Accurate taxi demand forecasting could help the taxi management platform to allocate taxi resources in advance, alleviate traffic congestion, and reduce passenger waiting time. Thus, more efforts in industrial and academic circles have been directed towards the cities’ taxi service demand prediction (CTSDP). However, the complex nonlinear spatio-temporal relationship in demand data makes it challenging to construct an accurate forecasting model. There remain challenges in perceiving the micro spatial characteristics and the macro periodicity characteristics from cities’ taxi service demand data. What’s more, the existing methods are significantly insufficient for exploring the potential multi-time patterns from these demand data. To meet the above challenges, and also stimulated by the human perception mechanism, we propose a Multi-Sensory Stimulus Attention (MSSA) model for CTSDP. Specifically, the MSSA model integrates a detail perception attention and a stimulus variety attention for capturing the micro and macro characteristics from massive historical demand data, respectively. The multiple time resolution modules are employed to capture multiple potential spatio-temporal periodic features from massive historical demand data. Extensive experiments on the yellow taxi trip records data in Manhattan show that the MSSA model outperforms the state-of-the-art baselines.