DeepRoute+: Modeling Couriers’ Spatial-temporal Behaviors and Decision Preferences for Package Pick-up Route Prediction

Over 10 billion packages are picked up every day in China. A fundamental task raised in the emerging intelligent logistics systems is the couriers’ package pick-up route prediction, which is beneficial for package dispatching, arrival-time estimation and overdue-risk evaluation, by leveraging the predicted routes to improve those downstream tasks. In the package pick-up scene, the decision-making of a courier is affected by strict spatial-temporal constraints (e.g., package location, promised pick-up time, current time, and courier’s current location). Furthermore, couriers have different decision preferences on various factors (e.g., time factor, distance factor, and balance of both), based on their own perception of the environments and work experience. In this article, we propose a novel model, named DeepRoute+, to predict couriers’ future package pick-up routes according to the couriers’ decision experience and preference learned from the historical behaviors. Specifically, DeepRoute+ consists of three layers: (1) The representation layer produces experience- and preference-aware representations for the unpicked-up packages, in which a decision preference module can dynamically adjust the importance of factors that affects the courier’s decision under the current situation. (2) The transformer encoder layer encodes the representations of packages while considering the spatial-temporal correlations among them. (3) The attention-based decoder layer uses the attention mechanism to generate the whole pick-up route recurrently. Experiments on a real-world logistics dataset demonstrate the state-of-the-art performance of our model.

Critical-like bistable dynamics in the resting-state human brain

Brain activity exhibits scale-free avalanche dynamics and power-law long-range temporal correlations (LRTCs) across the nervous system. This has been thought to reflect "brain criticality", i.e., brains operating near a critical phase transition between disorder and excessive order. Neuronal activity is, however, metabolically costly and may be constrained by activity-limiting mechanisms and resource depletion, which could make the phase transition discontinuous and bistable. Observations of bistability in awake human brain activity have nonetheless remained scarce and its functional significance unclear. First, using computational modelling where bistable synchronization dynamics emerged through local positive feedback, we found bistability to occur exclusively in a regime of critical-like dynamics. We then assessed bistability in vivo with resting-state magnetoencephalography and stereo-encephalography. Bistability was a robust characteristic of cortical oscillations throughout frequency bands from δ (3–7 Hz) to high-γ (100–225 Hz). As predicted by modelling, bistability and LRTCs were positively correlated. Importantly, while moderate levels of bistability were positively correlated with executive functioning, excessive bistability was associated with epileptic pathophysiology and predictive of local epileptogenicity. Critical bistability is thus a salient feature of spontaneous human brain dynamics in awake resting-state and is both functionally and clinically significant. These findings expand the framework of brain criticality and show that critical-like neuronal dynamics in vivo involves both continuous and discontinuous phase transitions in a frequency-, neuroanatomy-, and state-dependent manner.

Passenger Flow Prediction Using Smart Card Data from Connected Bus System Based on Interpretable XGBoost

Bus passenger flow prediction is a critical component of advanced transportation information system for public traffic management, control, and dispatch. With the development of artificial intelligence, many previous studies attempted to apply machine learning models to extract comprehensive correlations from transit networks to improve passenger flow prediction accuracy, given that the variety and volume of traffic data have been easily obtained. The passenger flow on a station is highly affected by various factors such as the previous time step, peak hours or nonpeak hours, and extracting the key features from the data is essential for a passenger flow prediction model. Although the neural networks, k -nearest neighbor, and some deep learning models have been adopted to mine the temporal correlations of the passenger flow data, the lack of interpretability of the influenced variables is still a big problem. Classical tree-based models can mine the correlations between variables and rank the importance of each variable. In this study, we presented a method to extract passenger flow of different routes on the station and implemented a XGBoost model to find the contributions of variables to the prediction of passenger flow. Comparing to benchmark models, the proposed model can reach state-of-the-art prediction accuracy and computational efficiency on the real-world dataset. Moreover, the XGBoost model can interpret the predicted results. It can be seen that period is the most important variable for the passenger flow prediction, and so the management of buses during peak hours should be improved.

Multi-temporal runoff-sediment discharge relationships

To understand the runoff-sediment discharge relationship , this study examined the annual runoff and sediment discharge data obtained from the Tangnaihai hydrometric station. The data were decomposed into multiple time scales through Complete Ensemble Empirical Mode Decomposition with adaptive noise (CEEMDAN). Furthermore, double cumulative curves were plotted and the cointegration theory was employed to analyze the microscopic and macroscopic multi-temporal correlations between the runoff and the sediment discharge and their detailed evolution.

Universal spectrum for short period (days) variability in atmospheric total ozone

Long-range spatio-temporal correlations manifested as the self-similar fractal geometry to the spatial pattern concomitant with inverse power law form for the power spectrum of temporal fluctuations are ubiquitous to real world dynamical systems and are recently identified as signatures of self-organized criticality Self-organised criticality in atmospheric flows is exhibited as the fractal geometry 10 the global cloud cover pattern and the inverse power law form for the atmospheric eddy energy spectrum, In this paper, a recently developed cell dynamical system model for atmospheric flows is summarized. The model predicts inverse power law form of the statistical normal distribution for atmospheric eddy energy spectrum as a natural consequence of quantum-like mechanics governing atmospheric flows extending up to stratospheric levels and above, Model Predictions are in agreement with continuous periodogram analyses of atmospheric total ozone. Atmospheric total ozone variability (in days) exhibits the temporal signature of self-organized criticality, namely, inverse power law form for the power spectrum. Further, the long-range temporal correlations implicit to self-organized criticality can be quantified in terms of the universal characteristics of the normal distribution. Therefore the total pattern of fluctuations of total ozone over a period of time is predictable.

Time-resolved smoothness of distributed brain activity tracks conscious states and unifies emergent neural phenomena

Much of systems neuroscience posits that emergent neural phenomena underpin important aspects of brain function. Studies in the field variously emphasize the importance of distinct emergent phenomena, including weakly stable dynamics, arrhythmic 1/f activity, long-range temporal correlations, and scale-free avalanche statistics. Few studies, however, have sought to reconcile these often abstract phenomena with interpretable properties of neural activity. Here, we developed a method to efficiently and unbiasedly generate model data constrained by interpretable empirical features in long neurophysiological recordings. We used this method to ground several major emergent neural phenomena to time-resolved smoothness, the correlation of distributed brain activity between adjacent timepoints. We first found that in electrocorticography recordings, time-resolved smoothness closely tracked transitions between conscious and anesthetized states. We then showed that a minimal model constrained by time-resolved smoothness, variance, and mean, captured dynamical and statistical emergent neural phenomena across modalities and species. Our results thus decouple major emergent neural phenomena from network mechanisms of brain function, and instead couple these phenomena to spatially nonspecific, time-resolved changes of brain activity. These results anchor several theoretical frameworks to a single interpretable property of the neurophysiological signal and, in this way, ultimately help bridge abstract theories of brain function with observed properties of brain activity.

Change of Persistence in European Electricity Spot Prices

The European Power Exchange has introduced day-ahead auctions and continuous trading spot markets to facilitate the insertion of renewable electricity. These markets are designed to balance excess or lack of power in short time periods, which leads to a large stochastic variability of the electricity prices. Furthermore, the different markets show different stochastic memory in their electricity price time series, which seem to be the cause for the large volatility. In particular, we show the antithetical temporal correlation in the intraday 15 minutes spot markets in comparison to the day-ahead hourly market. We contrast the results from Detrended Fluctuation Analysis (DFA) to a new method based on the Kramers–Moyal equation in scale. For very short term (< 12 hours), all price time series show positive temporal correlations (Hurst exponent H > 0.5) except for the intraday 15 minute market, which shows strong negative correlations (H < 0.5). For longer term periods covering up to two days, all price time series are anti-correlated (H < 0.5).

Proposal and Investigation of a Convolutional and LSTM Neural Network for the Cost-Aware Resource Prediction in Softwarized Networks

Resource prediction algorithms have been recently proposed in Network Function Virtualization architectures. A prediction-based resource allocation is characterized by higher operation costs due to: (i) Resource underestimate that leads to quality of service degradation; (ii) used cloud resource over allocation when a resource overestimate occurs. To reduce such a cost, we propose a cost-aware prediction algorithm able to minimize the sum of the two cost components. The proposed prediction solution is based on a convolutional and Long Short Term Memory neural network to handle the spatial and temporal correlations of the need processing capacities. We compare in a real network and traffic scenario the proposed technique to a traditional one in which the aim is to exactly predict the needed processing capacity. We show how the proposed solution allows for cost advantages in the order of 20%.