Public bikes have emerged as a significant mode of transportation for commuters. However, public bikes can be damaged to varying degrees as use frequency increases. The stations occupied by damaged bikes are of low usability, which may result in fewer users and a waste of resources. In this paper, we investigate a global scheduling solution for a bike system with unbalanced life-value distribution and design a scheduling approach. Firstly, we employ the Weibull model to estimate bike life to quantify use load. Moreover, we design a clustering algorithm to partition station regions. We also design a demand prediction model named ST-SAGCN to capture dynamic spatial correlations and spatio-temporal correlations simultaneously. We also propose a scheduling method that meets station demand as much as possible to balance station use and bike-life distribution. We conduct experiments on two public bike datasets covering different time spans in New York and Washington, and compare the bike-life distribution after scheduling with that of real-world actual conditions. The experimental results attest to the effectiveness of our approach to balancing bike use-load. The code is publicly available athttps://github.com/Zayn-Tang/BikeSchedulingStrategy.

Abstract Taxi demand forecasting from origin to destination (OD) is an important component in managing public transportation needs on a city‐wide scale. Accurate taxi demand forecasting may provide several benefits, including economic and traffic flow optimization. However, due to complicated spatial–temporal connections and irregular distant locations, predicting taxi demand becomes difficult. To address these issues, we proposed a novel architecture of multi‐module spatial–temporal networks to collectively predict city‐wide OD taxi demand. In our work to deal with adjacent areas in a city, we employed the 3D convolutional neural networks to extract the spatial–temporal dependencies and learn the OD taxi demand pattern. To handle remote areas, we created an attention‐based auto encoder‐decoder, in which the input of the set of convolutional layers creates a feature matrix. The feature matrix embeds the spatial–temporal correlation jointly and passing through the encoder layer. We encode the spatial–temporal features with the help of pooling layer, then flatten layer used the back‐propagation method to decode the weight matrix. We apply the normalization function to determine the demand pattern influences. The influence vector we compute with the Euclidean distance formula to determine the similarity of all distant regions. Finally, we use the attention mechanism to calculate the attention weight score for each region that its neighbour impacted. Then we used long short‐term memory, we captured the significant relationship of spatial–temporal dependencies with external factors. We train our model simultaneously to forecast city‐wide taxi services OD. We have performed comprehensive experiments and large‐scale dataset comparisons that reveal the taxi demand prediction problem.

Location prediction has attracted much attention due to its important role in many location-based services, including taxi services, route navigation, traffic planning, and location-based advertisements. Traditional methods only use spatial-temporal trajectory data to predict where a user will go next. The divorce of semantic knowledge from the spatial-temporal one inhibits our better understanding of users’ activities. Inspired by the architecture of Long Short Term Memory (LSTM), we design ST-LSTM, which draws on semantic trajectories to predict future locations. Semantic data add a new dimension to our study, increasing the accuracy of prediction. Since semantic trajectories are sparser than the spatial-temporal ones, we propose a strategic filling algorithm to solve this problem. In addition, as the prediction is based on the historical trajectories of users, the cold-start problem arises. We build a new virtual social network for users to resolve the issue. Experiments on two real-world datasets show that the performance of our method is superior to those of the baselines.

Traffic speed inference enables many applications that are essential for everyday life. Most traffic-prediction approaches assume that a constant number of sensors are deployed on the roads, whether they are either stationary loop detectors or vehicles equipped with Global Positioning System (GPS) tracking devices. The static nature of those fixtures limits their ability to adapt to scenarios that are more dynamic. Rather than relying on several fixed sensors to detect changes and infer traffic, we use crowdsourcing to judiciously select individuals and then make predictions. Our solution consists of three core components: dynamic seed selection, regional cluster building and ensemble traffic prediction. In the first phase, we employ Efficient Transition Probability (ETP) to evaluate candidate seed sets. Road clusters are then formed using hierarchical clustering that is tweaked by a dynamic programming technique. This method assesses the eccentricity of every cluster to bond every road within each cluster more closely. Subsequently, we develop an ensemble-learning strategy in conjunction with Lasso regression to forecast traffic. The strength of our ensemble approach is its ability to manage absent seeds, a condition that has never been investigated, to our knowledge. Substantial experimental evaluation indicates that our claim of dynamic updates is valid and effective. Our solution outperforms the state-of-the-art techniques by a wide margin, in terms of prediction accuracy.

Destination prediction is an essential task in various mobile applications and up to now many methods have been proposed. However, existing methods usually suffer from the problems of heavy computational burden, data sparsity, and low coverage. Therefore, a novel approach named DestPD is proposed to tackle the aforementioned problems. Differing from an earlier approach that only considers the starting and current location of a partial trip, DestPD first determines the most likely future location and then predicts the destination. It comprises two phases, the offline training and the online prediction. During the offline training, transition probabilities between two locations are obtained via Markov transition matrix multiplication. In order to improve the efficiency of matrix multiplication, we propose two data constructs, Efficient Transition Probability (ETP) and Transition Probabilities with Detours (TPD). They are capable of pinpointing the minimum amount of needed computation. During the online prediction, we design Obligatory Update Point (OUP) and Transition Affected Area (TAA) to accelerate the frequent update of ETP and TPD for recomputing the transition probabilities. Moreover, a new future trajectory prediction approach is devised. It captures the most recent movement based on a query trajectory. It consists of two components: similarity finding through Best Path Notation (BPN) and best node selection. Our novel BPN similarity finding scheme keeps track of the nodes that induces inefficiency and then finds similarity fast based on these nodes. It is particularly suitable for trajectories with overlapping segments. Finally, the destination is predicted by combining transition probabilities and the most probable future location through Bayesian reasoning. The DestPD method is proved to achieve one order of cut in both time and space complexity. Furthermore, the experimental results on real-world and synthetic datasets have shown that DestPD consistently surpasses the state-of-the-art methods in terms of both efficiency (approximately over 100 times faster) and accuracy.