Search

Traffic forecasting is a critical task to extract values from cyber-physical infrastructures, which is the backbone of smart transportation. However owing to external contexts, the dynamics at each sensor are unique. For example, the afternoon peaks at sensors near schools are more likely to occur earlier than those near residential areas. In this paper, we first analyze real-world traffic data to show that each sensor has a unique dynamic. Further analysis also shows that each pair of sensors also has a unique dynamic. Then, we explore how node embedding learns the unique dynamics at every sensor location. Next, we propose a novel module called Spatial Graph Transformers (SGT) where we use node embedding to leverage the self-attention mechanism to ensure that the information flow between two sensors is adaptive with respect to the unique dynamic of each pair. Finally, we present Graph Self-attention WaveNet (G-SWaN) to address the complex, non-linear spatiotemporal traffic dynamics. Through empirical experiments on four real-world, open datasets, we show that the proposed method achieves superior performance on both traffic speed and flow forecasting. Code is available at: https://github.com/aprbw/G-SWaN, 20 pages, IoTDI 2023; Correction on Fig. 4

Understanding human activity patterns in cities enables a more efficient and sustainable energy, transport, and resource planning. In this invited talk, after laying out the background on spatio-temporal representation, I will present our unsupervised approaches to handle large-scale mutivariate sensor data from heterogeneous sources, prior to modelling them further with the rich contextual signals obtained from the environment. I will also present several spatio-temporal prediction and recommendation problems, leveraging graph-based enrichment and embedding techniques, with applications in continuous trajectory prediction, visitor intent profiling, and urban flow forecasting.

Change Point Detection (CPD) methods identify the times associated with changes in the trends and properties of time series data in order to describe the underlying behaviour of the system. For instance, detecting the changes and anomalies associated with web service usage, application usage or human behaviour can provide valuable insights for downstream modelling tasks. We propose a novel approach for self-supervised Time Series Change Point detection method based onContrastivePredictive coding (TS-CP^2). TS-CP^2 is the first approach to employ a contrastive learning strategy for CPD by learning an embedded representation that separates pairs of embeddings of time adjacent intervals from pairs of interval embeddings separated across time. Through extensive experiments on three diverse, widely used time series datasets, we demonstrate that our method outperforms five state-of-the-art CPD methods, which include unsupervised and semi-supervisedapproaches. TS-CP^2 is shown to improve the performance of methods that use either handcrafted statistical or temporal features by 79.4% and deep learning-based methods by 17.0% with respect to the F1-score averaged across the three datasets., Accepted at The WEB Conference 2021 (WWW'21)

Network alignment is useful for multiple applications that require increasingly large graphs to be processed. Existing research approaches this as an optimization problem or computes the similarity based on node representations. However, the process of aligning every pair of nodes between relatively large networks is time-consuming and resource-intensive. In this paper, we propose a framework, called G-CREWE (Graph CompREssion With Embedding) to solve the network alignment problem. G-CREWE uses node embeddings to align the networks on two levels of resolution, a fine resolution given by the original network and a coarse resolution given by a compressed version, to achieve an efficient and effective network alignment. The framework first extracts node features and learns the node embedding via a Graph Convolutional Network (GCN). Then, node embedding helps to guide the process of graph compression and finally improve the alignment performance. As part of G-CREWE, we also propose a new compression mechanism called MERGE (Minimum dEgRee neiGhbors comprEssion) to reduce the size of the input networks while preserving the consistency in their topological structure. Experiments on all real networks show that our method is more than twice as fast as the most competitive existing methods while maintaining high accuracy., 10 pages, accepted at the 29th ACM International Conference onInformation and Knowledge Management (CIKM 20)

We introduce the novel research problem of task recognition in daily life. We recognize tasks such as project management, planning, meal-breaks, communication, documentation, and family care. We capture Cyber, Physical, and Social (CPS) activities of 17 participants over four weeks using device-based sensing, app activity logging, and an experience sampling methodology. Our cohort includes students, casual workers, and professionals, forming the first real-world context-rich task behaviour dataset. We model CPS activities across different task categories, results highlight the importance of considering the CPS feature sets in modelling, especially work-related tasks.

Obstructive sleep apnea (OSA) is a serious disorder in which people repeatedly stop breathing during sleep. This paper presents the Technology Integrated Apnea Respiration Analyser (TIARA), an innovative wearable device as an alternative to polysomnography (PSG) that has the potential to enable cost-effective and repeated overnight tests of OSA even in a home environment. PSG normally collects bio-signals such as brain waves, oxygen saturation, heart rate in a controlled environment (e.g. sleep lab) to diagnose sleep disorders. Here we present the design and development of the TIARA device and demonstrate its potential to perform as well as PSG at differentiating sleep and wake states and different phases of sleep, based on machine learning models.

Intelligent assistants can serve many purposes, including entertainment (e.g. playing music), home automation, and task management (e.g. timers, reminders). The role of these assistants is evolving to also support people engaged in work tasks, in workplaces and beyond. To design truly useful intelligent assistants for work, it is important to better understand the work tasks that people are performing. Based on a survey of 401 respondents' daily tasks and activities in a work setting, we present a classification of work-related tasks, and analyze their key characteristics, including the frequency of their self-reported tasks, the environment in which they undertake the tasks, and which, if any, electronic devices are used. We also investigate the cyber, physical, and social aspects of tasks. Finally, we reflect on how intelligent assistants could influence and help people in a work environment to complete their tasks, and synthesize our findings to provide insight on the future of intelligent assistants in support of amplifying personal productivity.

The developments in sensing modalities and computing platforms enable many new sensing technologies and data sources for monitoring occupant presence and actions. The wealth of data opens new opportunities for extracting knowledge through data-driven modeling of occupant presence and actions. In particular, the many opportunities with machine learning techniques including supervised and unsupervised learning for classification, regression and clustering problems. Utilizing these opportunities creates new models and information relevant for generating new knowledge on multi-aspect environmental exposure, building interfaces, human behaviour, occupant-centric building design and operation. Subtask 2 of the new IEA EBC Annex 79 is addressing these opportunities and is inviting researchers and practitioners to participate. The developed data-driven models can, among others, be applied for example for calculating new schedules or models based on the actual conditions observed in buildings, data-driven analysis of the performance design versus the built, model predictive controls for buildings and fault detection and diagnostics.

This paper investigates the Cyber-Physical behavior of a user in a large indoor shopping center by leveraging anonymized (opt in) Wi-Fi association and browsing logs recorded by the center operators. Our analysis shows that many users exhibit high correlation between their cyber activities and physical context. To find this correlation, we propose a mechanism to semantically label a physical space with rich categorical information from Wikipedia concepts and compute a contextual similarity that represents a customer's activities with the mall context. We further show the use of cyber-physical contextual similarity in two different applications: user behavior classification and future location prediction. The experimental results demonstrate that the users' contextual similarity significantly improves the accuracy of such applications.

Location-Based Social Networks (LBSN) provides unprecedented opportunities to tackle various social problems. In this study, we identify a number of crime-prediction-specific dynamic features which, for the first time, explore crime risk factors implicitly associated with the visitors. The reliable correlations between the proposed dynamic features and crime event occurrences have been observed. The evaluations on large real world data sets verify that the crime prediction performance can be notably improved with the inclusion of proposed crime-prediction-specific dynamic features.

In this paper, we focus on designing and developing ProMETheus, an intelligent system for meeting minutes generated from audio data. The first task in ProMETheus is to recognize the speakers from noisy audio data. Speaker recognition algorithm is used to automatically identify who is speaking according to the speech in an audio data. Naturally, speech recognition will transcribe speakers' audio to text so that ProMETheus can generate the complete meeting text with speakers' name chronologically. In order to show the subject of the meeting and the agreed action, we use text summarization algorithm that can extract meaningful key phrases and summary sentences from the complete meeting text. In addition, sentiment analysis for meeting text of different speakers can make the agreed action more humane due to calculating the relevance score of each course by the sentiment and attitude in text tone. The ProMETheus is capable of accurately summarizing the meeting and analyzing the agreed action. Our robust system is evaluated on a real-world audio meeting dataset that involves multiple speakers in each meeting session.

In this paper, we focus on simultaneous inference of transportation modes and human activities in daily life via modelling and inference from multivariate time series data, which are streamed from off-the-shelf mobile sensors (e.g. embedded in smartphones) in real-world dynamic environments. The transportation mode will be inferred from the structured hierarchical contexts associated with human activities. Through our mobile context recognition system, an accurate and robust solution can be obtained to infer transportation mode, human activity and their associated contexts (e.g. whether the user is in moving or stationary environment) simultaneously. There are many challenges in analysing and modelling human mobility patterns within urban areas due to the ever-changing environments of mobile users. For instance, a user could stay at a particular location and then travel to various destinations depending on the tasks they carry within a day. Consequently, there is a need to reduce the reliance on location-based sensors (e.g. GPS), since they consume a significant amount of energy on smart devices, for the purpose of intelligent mobile sensing (i.e. automatic inference of transportation mode, human activity and associated contexts). Nevertheless, our system is capable of outperforming the simplistic approach that only considers independent classifications of multiple context label sets on data streamed from low-energy sensors.

The global urbanization imposes unprecedented pressure on urban infrastructure and public resources. The population explosion has made it challenging to satisfy the daily needs of urban residents. 'Smart City' is a solution that utilizes different types of data collection sensors to help manage assets and resources intelligently and more efficiently. Under the Smart City umbrella, the primary research initiative in improving the efficiency of car-hailing services is to predict the citywide passenger demand to address the imbalance between the demand and supply. However, predicting the passenger demand requires analysis on various data such as historical passenger demand, crowd outflow, and weather information, and it remains challenging to discover the latent relationships among these data. To address this challenge, we propose to improve the passenger demand prediction via learning the salient spatial-temporal dynamics within a reinforcement learning framework. Our model employs an information selection mechanism to focus on the most distinctive data in historical observations. This mechanism can automatically adjust the information zone according to the prediction performance to find the optimal choice. It also ensures the prediction model to take full advantage of the available data by introducing the positive and excluding the negative correlations. We have conducted experiments on a large-scale real-world dataset that covers 1.5 million people in a major city in China. The results show our model outperforms state-of-the-art and a series of baselines by a large margin.

This paper addresses the problem of taxi-passenger queue context prediction using neighborhood based methods. We capture the taxi drivers' knowledge based on how they move in terms of temporal driver-knowledge deviation (TDKD). Then a TDKD-aided feature importance scheme is introduced for neighborhood based queue context prediction. We apply our proposed scheme to predict different queue contexts at a busy international airport in New York. We argue that the incorporation of taxi drivers' knowledge for calculating feature importance significantly improves the quality of selected neighborhood, thus boosting the prediction accuracy. The experimental results demonstrate the effectiveness of our proposed TDKD-aided feature importance scheme for neighborhood based taxi-passenger queue context prediction.

Understanding human mobility is the key problem in many applications such as location-based services and recommendation systems. The mobility of a smartphone user can be modeled by a movement graph, in which the nodes represent locations and the edges are distances or traveling times between the locations. However, the resulting graph would be too big to be stored and queried on resource-devices such as smartphones. In this paper, we deploy a state-of-the-art graph summarization method to produce an abstract (coarse) graph easy to be processed and queried. After summarization, the movement graph becomes smaller resulting in a reduction in the required time and storage to deploy graph algorithms. We specifically investigate the effect of summarization on two algorithms related to human mobility mining: location prediction and similarity mining. The location prediction algorithm on the coarse graph causes coarse-grain results. Regarding computing the similarity, summarization reduces the computational cost but at the same time increases the uncertainty of the results. We show that the trade-off between accuracy, storage space and speed can be controlled by the compression ratio. As an illustration, if the size of the graph is reduced to half, the similarity algorithm becomes 4 times faster while the correlation between similarities of coarse and original graphs is 0.98.

Human occupancy counting is crucial for both space utilisation and building energy optimisation. In the current article, we present a semi-supervised domain adaptation method for carbon dioxide - Human Occupancy Counter (DA-HOC), a robust way to estimate the number of people within in one room by using data from a carbon dioxide sensor. In our previous work, the proposed Seasonal Decomposition for Human Occupancy Counting (SD-HOC) model can accurately predict the number of individuals when the training and labelled data are adequately available. DA-HOC is able to predict the number of occupancy with minimal training data, as little as one-day data. DA-HOC accurately predicts indoor human occupancy for a large room using a model trained from a small room and adapted to the larger room. We evaluate DA-HOC with two baseline methods - support vector regression technique and SD-HOC model. The results demonstrate that DA-HOC's performance is better by 12.29% in comparison to SVR and 10.14% in comparison to SD-HOC.

In this paper, we focus on developing a model and system for predicting the city foot traffic. We utilise historical records of pedestrian counts captured with thermal and laser-based sensors installed at multiple locations throughout the city. A robust prediction system is proposed to cope with various temporal foot traffic patterns. The empirical evaluation of our experiment shows that the proposed ARIMA model is effective in modelling both weekdays and weekend patterns, outperforming other state-of-art models for short-term prediction of pedestrian counts. The model is capable of accurately predicting pedestrian numbers up to 16 days in advance, on multiple look-ahead times. Our system is evaluated with a real-world sensor dataset supplied by the City of Melbourne.

Understanding and predicting human mobility is a key problem in different applications. Existing works on human mobility prediction mainly focus on the prediction of the next location (or a set of locations) that will be visited by the user in a specified time. These methods do not take into account the locations sequence and the departure times. In this research, given the historical data and the user's trajectory in the first part of the current day (e.g trajectory in the morning), we introduce the concept of full trajectory prediction for the rest of the day (e.g. prediction of the trajectory in the afternoon). We emphasize that the full trajectory includes the sequence of the locations, the staying times, and the departure times. The proposed method has been examined on either labeled trajectories and geographical trajectories and the results show the effectiveness of the proposed method.

Multi-activity recognition in the urban environment is a challenging task. This is largely attributed to the influence of urban dynamics, the variety of the label sets, and the heterogeneous nature of sensor data that arrive irregularly and at different rates. One of the first tasks in multi-activity recognition is temporal segmentation. A common temporal segmentation method is the sliding window approach with a fixed window size, which is widely used for single activity recognition. In order to recognise multiple activities from heterogeneous sensor streams, we propose a new time windowing technique that can optimally extract segments with multiple activity labels. The mixture of activity labels causes the impurity in the corresponding temporal segment. Hence, larger window size imposes higher impurity in temporal segments while increasing class separability. In addition, the combination of labels from multiple activity label sets (i.e. number of unique multi-activity) may decrease as impurity increases. Naturally, these factors will affect the performance of classification task. In our proposed technique, the optimal window size is found by gaining the balance between minimising impurity and maximising class separability in temporal segments. As a result, it accelerates the learning process for recognising multiple activities (such as higher level and atomic human activities under different environment contexts) in comparison to laborious tasks of sensitivity analysis. The evaluation was validated by experiments on a real-world dataset for recognising multiple human activities in a smart environment.