Search

Abstract Accurate demand forecasting has become increasingly necessary in the burgeoning field of free‐floating micro‐mobility systems. However, for model training, the service area must be divided into specific areal units, which often involves grid‐based methods. Although these methods are feasible and provide a uniform area division, they are highly susceptible to the Modifiable Areal Unit Problem (MAUP), which is a critical issue in spatial data analysis. Although MAUP can adversely affect predictive model learning, studies addressing this issue are scarce. Therefore, a novel base areal unit generation algorithm is proposed that employs a clustering approach to enhance the prediction accuracy in free‐floating micro‐mobility system demand. The method identifies suitable base areal units by merging smaller ones while considering the similarities in temporal usage patterns and distances between different areas, mitigating the impact of MAUP during model learning. The approach was evaluated using shared e‐scooter data from two cities, Kansas City and Minneapolis, and it was compared to the traditional grid method. The findings indicate that the proposed framework generally improves prediction performance within the newly defined areal units.

Accurate demand forecasting has become increasingly necessary in the burgeoning field of free‐floating micro‐mobility systems. However, for model training, the service area must be divided into specific areal units, which often involves grid‐based methods. Although these methods are feasible and provide a uniform area division, they are highly susceptible to the Modifiable Areal Unit Problem (MAUP), which is a critical issue in spatial data analysis. Although MAUP can adversely affect predictive model learning, studies addressing this issue are scarce. Therefore, a novel base areal unit generation algorithm is proposed that employs a clustering approach to enhance the prediction accuracy in free‐floating micro‐mobility system demand. The method identifies suitable base areal units by merging smaller ones while considering the similarities in temporal usage patterns and distances between different areas, mitigating the impact of MAUP during model learning. The approach was evaluated using shared e‐scooter data from two cities, Kansas City and Minneapolis, and it was compared to the traditional grid method. The findings indicate that the proposed framework generally improves prediction performance within the newly defined areal units. [ABSTRACT FROM AUTHOR]

Abstract Online car‐hailing, with its advantages of convenience and efficiency, has quickly become popular among tourists, playing a crucial role in the accessibility of scenic spots. Due to the particularities of tourist travel behaviour and the complexity of travel supply and demand around scenic spots, research on car‐hailing tourists is relatively lacking at this stage. Based on multi‐source data, this study couples the identifying of travel characteristics, by introducing the concept of service dependency degree, with a Bayesian optimization–long short‐term memory–convolutional neural network (BO‐LSTM‐CNN) method to conduct multi‐task online car‐hailing demand forecasting. The evaluation of the dependency degree primarily encompasses the establishment of evaluation indices and the application of the entropy weight method and natural breakpoint method. The BO‐LSTM‐CNN model utilizes Bayesian optimization for hyperparameter tuning, LSTM for temporal variable processing, and CNN for the fusion of multi‐source information related to weather, space, and online car‐hailing attributes. Extracting online car‐hailing tourist travel orders based on spatial–temporal constraints, the proposed methods are applied to 72 scenic spots in Beijing, China. According to their dependency degree, Beijing's scenic spots are categorized into three levels of dependency on online car‐hailing services, from high to low. The outstanding forecasting efficacy of the proposed model for various scenic spots is verified through comparison tests with several benchmark models. Consequent to these findings, mobility service improvement strategies are specifically proposed for each class of scenic spots, which can provide valuable insights for the relevant tourism traffic management personnel.

Online car‐hailing, with its advantages of convenience and efficiency, has quickly become popular among tourists, playing a crucial role in the accessibility of scenic spots. Due to the particularities of tourist travel behaviour and the complexity of travel supply and demand around scenic spots, research on car‐hailing tourists is relatively lacking at this stage. Based on multi‐source data, this study couples the identifying of travel characteristics, by introducing the concept of service dependency degree, with a Bayesian optimization–long short‐term memory–convolutional neural network (BO‐LSTM‐CNN) method to conduct multi‐task online car‐hailing demand forecasting. The evaluation of the dependency degree primarily encompasses the establishment of evaluation indices and the application of the entropy weight method and natural breakpoint method. The BO‐LSTM‐CNN model utilizes Bayesian optimization for hyperparameter tuning, LSTM for temporal variable processing, and CNN for the fusion of multi‐source information related to weather, space, and online car‐hailing attributes. Extracting online car‐hailing tourist travel orders based on spatial–temporal constraints, the proposed methods are applied to 72 scenic spots in Beijing, China. According to their dependency degree, Beijing's scenic spots are categorized into three levels of dependency on online car‐hailing services, from high to low. The outstanding forecasting efficacy of the proposed model for various scenic spots is verified through comparison tests with several benchmark models. Consequent to these findings, mobility service improvement strategies are specifically proposed for each class of scenic spots, which can provide valuable insights for the relevant tourism traffic management personnel. [ABSTRACT FROM AUTHOR]

Abstract Accurate traffic flow forecasting is a critical component in intelligent transportation systems. However, most of the existing traffic flow prediction algorithms only consider the prediction under normal conditions, but not the influence of weather attributes on the prediction results. This study applies a hybrid deep learning model based on multi feature fusion to predict traffic flow considering weather conditions. A comparison with other representative models validates that the proposed spatial‐temporal fusion graph convolutional network (STFGCN) can achieve better performance.

Abstract Urban traffic demand forecasting is an important component in the implementation of intelligent transport systems (ITS). Urban traffic demand data is a spatiotemporal data, describing the amount of traffic demand generated by different areas or stations within a city along the time dimension. Although there has been considerable research work, researchers still face several challenges in predicting accurately, including the capture of hidden features in the temporal dimension of such spatiotemporal data, and the capture of dynamic dependent changes in the spatial dimension. These are even more difficult for long‐time series prediction tasks. This paper designs a multivariate temporal forecasting model specifically adapted to traffic demand to address these challenges, called the Gated Spatiotemporal Graph Attention Network (GSTGAT). GSTGAT is based on the Transformer framework and the whole model is used in an end‐to‐end manner. First of all, it uses the gated self‐attention to extract temporal features in the sequence. Secondly, graph attention is used to capture the spatial dependencies among different variables in the unstructured space. Finally, the use of gated recurrent units in combination with hidden spatial states is proposed to capture multiple levels of spatial dependencies. Experimental results on the taxi dataset in New York and the bicycle dataset in San Francisco Bay Area show that the authors’ proposed model outperforms other state‐of‐the‐art models and improves the prediction accuracy.

Abstract Cellular data is a sequence of base station‐interaction data that records user ID, timestamp, location area code (LAC), and cell identity (CID). With long observation periods, the data allows traffic planners to analyze coarse‐granularity user travel behaviours at low costs. However, utilizing cellular data for urban planning is not an easy task as the data lacks user socioeconomic attributes due to privacy issues. The data is also challenging to recognize user activity types. This paper proposed an activity‐based model (ABM) with skeleton schedule constraints for multi‐day cellular data. The model first infers the activity pattern and home location. Then it predicts start time, duration, and locations separately for primary and secondary activities. Next, the model infers the travel mode and path considering user multi‐day travel behaviour, path non‐linear coefficient, and transfers. Finally, a time adjustment module is proposed to avoid time conflicts in consecutive activities. The proposed activity‐based model is validated at activity, travel, and path levels. Results show that the proposed model can effectively predict activities and has much higher stability than existing ABMs based on travel surveys.

Accurate traffic flow forecasting is a critical component in intelligent transportation systems. However, most of the existing traffic flow prediction algorithms only consider the prediction under normal conditions, but not the influence of weather attributes on the prediction results. This study applies a hybrid deep learning model based on multi feature fusion to predict traffic flow considering weather conditions. A comparison with other representative models validates that the proposed spatial‐temporal fusion graph convolutional network (STFGCN) can achieve better performance. [ABSTRACT FROM AUTHOR]

Urban traffic demand forecasting is an important component in the implementation of intelligent transport systems (ITS). Urban traffic demand data is a spatiotemporal data, describing the amount of traffic demand generated by different areas or stations within a city along the time dimension. Although there has been considerable research work, researchers still face several challenges in predicting accurately, including the capture of hidden features in the temporal dimension of such spatiotemporal data, and the capture of dynamic dependent changes in the spatial dimension. These are even more difficult for long‐time series prediction tasks. This paper designs a multivariate temporal forecasting model specifically adapted to traffic demand to address these challenges, called the Gated Spatiotemporal Graph Attention Network (GSTGAT). GSTGAT is based on the Transformer framework and the whole model is used in an end‐to‐end manner. First of all, it uses the gated self‐attention to extract temporal features in the sequence. Secondly, graph attention is used to capture the spatial dependencies among different variables in the unstructured space. Finally, the use of gated recurrent units in combination with hidden spatial states is proposed to capture multiple levels of spatial dependencies. Experimental results on the taxi dataset in New York and the bicycle dataset in San Francisco Bay Area show that the authors' proposed model outperforms other state‐of‐the‐art models and improves the prediction accuracy. [ABSTRACT FROM AUTHOR]

Cellular data is a sequence of base station‐interaction data that records user ID, timestamp, location area code (LAC), and cell identity (CID). With long observation periods, the data allows traffic planners to analyze coarse‐granularity user travel behaviours at low costs. However, utilizing cellular data for urban planning is not an easy task as the data lacks user socioeconomic attributes due to privacy issues. The data is also challenging to recognize user activity types. This paper proposed an activity‐based model (ABM) with skeleton schedule constraints for multi‐day cellular data. The model first infers the activity pattern and home location. Then it predicts start time, duration, and locations separately for primary and secondary activities. Next, the model infers the travel mode and path considering user multi‐day travel behaviour, path non‐linear coefficient, and transfers. Finally, a time adjustment module is proposed to avoid time conflicts in consecutive activities. The proposed activity‐based model is validated at activity, travel, and path levels. Results show that the proposed model can effectively predict activities and has much higher stability than existing ABMs based on travel surveys. [ABSTRACT FROM AUTHOR]