Autonomous vehicles (AVs) are becoming integral to on-demand micro transit, offering the potential for safer, efficient, and sustainable transportation. However, AV deployment faces several challenges, including the lack of suitable roadways, varying travel conditions. Traditional routers prioritize speed and not reliability, leading to unpredictable operations and complications in planning. To address these, we introduce AVATAR, an autonomy-aware routing framework that prioritizes dependable, low-variance routes. Our approach encodes multiple objectives including road speed, speed variability, zoning areas, pedestrian encounters, and operator preferred roadways into edge-level routing engines. Objective optimized routes are generated, then scored using a multi-criteria decision-making process. User-configurable preference profiles, allow operators to define a balance between reliability and speed. AVATAR is a data-driven framework that supports both real-time AV operations and offline analysis, enabling transit operators to assess and refine routing strategies. Our experiments using real-world data from Silicon Valley, California, and Yokohama, Japan show that our approach significantly improves AV reliability and performance and advances the sustainable and scalable integration of AVs into future transportation networks.

In recent years, considerable progress has been made on improving the interpretability of machine learning models. This is essential, as complex deep learning models with millions of parameters produce state of the art performance, but it can be nearly impossible to explain their predictions. While various explainability techniques have achieved impressive results, nearly all of them assume each data instance to be independent and identically distributed (iid). This excludes relational models, such as Statistical Relational Learning (SRL), and the recently popular Graph Neural Networks (GNNs), resulting in few options to explain them. While there does exist work on explaining GNNs, GNN-Explainer, they assume access to the gradients of the model to learn explanations, which is restrictive in terms of its applicability across non-differentiable relational models and practicality. In this work, we develop RelEx, amodel-agnostic relational explainer to explain black-box relational models with only access to the outputs of the black-box. RelEx is able to explain any relational model, including SRL models and GNNs. We compare RelEx to the state-of-the-art relational explainer, GNN-Explainer, and relational extensions of iid explanation models and show that RelEx achieves comparable or better performance, while remaining model-agnostic.

Accurate wireless channel quality prediction over 4G LTE networks continues to be an important problem as future channel predictions are widely leveraged to meet the strict requirements of applications such as 360-degree video, ARlVR, and online games. The availability of large amounts of wireless channel data, the increase in computational power and the advancements in the field of machine learning provide us the opportunity to design learning-based approaches to address the channel quality prediction problem. In this paper, we design discriminative sequence-to-sequence probabilistic graphical models, specifically sparse Gaussian Conditional Random Fields (GCRF) models to accurately predict future channel quality variations in 4G LTE networks based on past channel quality data. In contrast to prior work that has primarily focused on designing parsimonious Markovian models or computationally-intensive deep learning models, the sparse GCRF models designed here provide superior performance while being highly interpretable and computationally efficient, thus making them an ideal choice for practical deployment. To validate the efficacy of our sparse GCRF model, we compare its performance (i.e., root mean squared error and mean absolute error) with i) linear regression and ii) ARIMA and iii) the state-of-the-art deep learning model on real-world 4G LTE channel quality data collected under varying levels of user mobility for two cellular operators and observe that the GCRF model provides significantly higher performance improvement.

In this paper, our goal is to model the aggregate mobility of individuals in a city by analyzing cellular network connections, and then leverage the designed mobility model to model and predict the number of COVID-19 infections in future. We analyze cellular network connections from 973 antennas for all users in the city of Rio de Janeiro from April 5, 2020 to July 2, 2020. We design a Markovian model that captures the mobility across municipalities. We then combine the transition probabilities of the Markov chain with the number of COVID-19 cases in a municipality during a particular week in the design of our mobility-aware COVID-19 case prediction models to predict the number of cases for the following week. Our experiments demonstrate that our mobility-aware models significantly out-perform a baseline mobility-agnostic linear regression model in terms of metrics such as Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE).

In this paper, our goal is to analyze and compare cellular network usage data from pre-lockdown, during lock-down, and post-lockdown phases surrounding the COVID-19 pandemic to understand and model human mobility patterns during the pandemic. To this end, we collect and analyze cellular network connections from 1400 antennas for all users in the city of Rio de Janeiro and its suburbs from March 1, 2020 to July 1, 2020. Our analysis reveals that the total number of cellular connections decreases to 78% during the lockdown phase and then increases to 85% of the pre-COVID era as the lockdown eases. We observe that user mobility starts increasing around 3 weeks before the end of lockdown, with the trend continuing into the post-lockdown period. We also design an interactive tool that showcases mobility patterns in different granularities and can help government officials take informed actions to control the spread of the disease.

In this paper, our goal is to leverage cellular network traffic data to model and forecast the number of COVID-19 infections in the future. To this end, we partner with one of the main cellular network providers in Brazil, TIM Brazil, and collect and analyze cellular network connections from 973 antennas for all users in the city of Rio de Janeiro and its suburbs. We develop a Markovian model that captures the mobility of individuals across municipalities of the city. The transition probabilities of the Markov chain are determined by analyzing user-level mobility events between antennas from the cellular network connectivity logs. We combine the aggregate mobility characteristics across municipalities as evidenced from the transition probabilities with the number of reported COVID-19 cases in a municipality during a particular week to design mobility-aware COVID-19 case prediction models that predict the number of cases for the following week. Our experiments demonstrate that our mobility-aware models significantly outperform a baseline mobility-agnostic linear regression model in terms of metrics such as Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE).

In this paper, our goal is to analyze and compare cellular network usage data from Rio de Janeiro from pre-lockdown, during lockdown, and post-lockdown phases surrounding the COVID-19 pandemic to understand and model human mobility patterns during the pandemic, and to evaluate the effect of lockdowns on mobility. Our analysis reveals that human mobility increases significantly even before lockdown restrictions are eased, with the trend continuing in the post-lockdown period. We also observe that the day of week has a significant impact on mobility of individuals, with the overall mobility on Fridays increasing over time possibly due to people self-relaxing restrictions and engaging in social activities on Friday evenings. We also design an interactive tool that showcases mobility patterns in different granularities and can potentially help people and government officials understand the mobility of individuals and the number of COVID-19 cases in a particular neighborhood.

In this paper, we collect data from Twitter and conduct a linguistic analysis of the user tweets to understand the social and economic disruption caused by the COVID-19 pandemic. To better appreciate peoples’ opinions and concerns with regards to the socio-economic conditions of addiction, mental health, unemployment and immigration, we collect data for a period of approximately 3 months in the beginning of the pandemic. We analyze the term and co-occurrence frequencies to identify the most commonly occurring words and bigrams in the discussion for each of the four categories. We conduct semantic role labeling to determine the action words in each category and then adopt a LSTM-based dependency parsing model to identify the main nouns linked with these action words. We then adopt a seeded topic modeling approach to automatically identify the main topics of discussion in each category. We finally conclude with a sentiment analysis of the tweets in each category to determine the overall sentiment associated with each category. Our fine-grained linguistic study unearths the difficulties experienced by the people (e.g., action verb need associated with nouns such as aid and assistance in the unemployment category). We also observe that the overall sentiment in the tweets is negative, driven by people experiencing the pains of job loss, deportation, and the difficulty in accessing programs and treatments related to addiction. Our analysis highlights the main challenges experienced by the people during the start of the COVID-19 crisis and lays the foundation for recognizing and developing the most pertinent public and social policies so as to minimize peoples’ suffering in case of a future pandemic.

Alcoholism, also known as Alcohol Use Disorder (AUD), is a serious problem affecting millions of people worldwide. Recovery from AUD is known to be challenging and often leads to relapse at various points after enrolling in a rehabilitation program such as Alcoholics Anonymous (AA). In this work, we present a structured and linguistic approach using hinge-loss Markov random fields (HL-MRFs) to understand recovery and relapse from AUD using social media data. We evaluate our models on AA-attending users extracted from: (i) the Twitter social network and predict recovery at two different points—90 days and 1 year after the user joins AA, respectively, and (ii) the Reddit AA recovery forums and predict whether the participating user is currently sober. The two datasets present two facets of the same underlying problem of understanding recovery and relapse in AUD users. We flesh out different characteristics in both these datasets: (i) In the Twitter dataset, we focus on the social aspect of the users and the relationship with recovery and relapse, and (ii) in the Reddit dataset, we focus on modeling the linguistic topics and dependency structure to understand users’ recovery journey. We design a unified modeling framework using HL-MRFs that takes the different characteristics of both these platforms into account. Our experiments reveal that our structured and linguistic approach is helpful in predicting recovery in users in both these datasets. We perform extensive quantitative analysis of different groups of features and dependencies among them in both datasets. The interpretable and intuitive nature of our models and analysis is helpful in making meaningful predictions and can potentially be helpful in identifying and preventing relapse early.

The mixed membership stochastic blockmodel (MMSB) is a popular framework for community detection and network generation. It learns a low-rank mixed membership representation for each node across communities by exploiting the underlying graph structure. MMSB assumes that the membership distributions of the nodes are independently drawn from a Dirichlet distribution, which limits its capability to model highly correlated graph structures that exist in real-world networks. In this paper, we present a flexible richly structured MMSB model, \textit{Struct-MMSB}, that uses a recently developed statistical relational learning model, hinge-loss Markov random fields (HL-MRFs), as a structured prior to model complex dependencies among node attributes, multi-relational links, and their relationship with mixed-membership distributions. Our model is specified using a probabilistic programming templating language that uses weighted first-order logic rules, which enhances the model's interpretability. Further, our model is capable of learning latent characteristics in real-world networks via meaningful latent variables encoded as a complex combination of observed features and membership distributions. We present an expectation-maximization based inference algorithm that learns latent variables and parameters iteratively, a scalable stochastic variation of the inference algorithm, and a method to learn the weights of HL-MRF structured priors. We evaluate our model on six datasets across three different types of networks and corresponding modeling scenarios and demonstrate that our models are able to achieve an improvement of 15\% on average in test log-likelihood and faster convergence when compared to state-of-the-art network models.

The development of fair machine learning models that effectively avert bias and discrimination is an important problem that has garnered attention in recent years. The necessity of encoding complex relational dependencies among the features and variables for competent predictions require the development of fair, yet expressive relational models. In this work, we introduce Fair-A3SL, a fairness-aware structure learning algorithm for learning relational structures, which incorporates fairness measures while learning relational graphical model structures. Our approach is versatile in being able to encode a wide range of fairness metrics such as statistical parity difference, overestimation, equalized odds, and equal opportunity, including recently proposed relational fairness measures. While existing approaches employ the fairness measures on pre-determined model structures post prediction, Fair-A3SL directly learns the structure while optimizing for the fairness measures and hence is able to remove any structural bias in the model. We demonstrate the effectiveness of our learned model structures when compared with the state-of-the-art fairness models quantitatively and qualitatively on datasets representing three different modeling scenarios: i) a relational dataset, ii) a recidivism prediction dataset widely used in studying discrimination, and iii) a recommender systems dataset. Our results show that Fair-A3SL can learn fair, yet interpretable and expressive structures capable of making accurate predictions.

Thoroughly understanding how energy consumption is disaggregated into individual appliances can help reduce household expenses, integrate renewable sources of energy, and lead to efficient use of energy. In this work, we propose a deep latent generative model based on variational recurrent neural networks (VRNNs) for energy disaggregation. Our model jointly disaggregates the aggregated energy signal into individual appliance signals, achieving superior performance when compared to the state-of-the-art models for energy disaggregation, yielding a 29% and 41% performance improvement on two energy datasets, respectively, without explicitly encoding temporal/contextual information or heuristics. Our model also achieves better prediction performance on lowpower appliances, paving the way for a more nuanced disaggregation model. The structured output prediction in our model helps in accurately discerning which appliance(s) contribute to the aggregated power consumption, thus providing a more useful and meaningful disaggregation model.

Accurately modeling and predicting wireless channel quality variations is essential for a number of networking applications such as scheduling and improved video streaming over 4G LTE networks and bit rate adaptation for improved performance in WiFi networks. In this paper, we propose an encoder-decoder based sequence-to-sequence deep learning model that is capable of predicting future wireless signal strength variations based on past signal strength data. We consider two different versions of the deep learning model; the first and second versions use LSTM and GRU as their basic cell structure, respectively. In contrast to prior work that is primarily focused on designing models for particular network settings, the deep learning model is highly adaptable and can predict future channel conditions for different networks, sampling rates, mobility patterns, and communication standards. We compare the performance (i.e., the root mean squared error of future predictions) of our model with respect to two baselines-i) auto-regression(1), and ii) linear regression for multiple networks and communication standards. In particular, we consider 4G LTE, WiFi, an industrial network operating in the 5.8 GHz range, Zigbee, and WiMAX networks operating under varying levels of user mobility and observe that the deep learning model provides significantly superior performance. Finally, we provide detailed discussion on key design decisions including insights into hyper-parameter tuning of the model.

Statistical relational models such as Markov logic networks (MLNs) and hinge-loss Markov random fields (HL-MRFs) are specified using templated weighted first-order logic clauses, leading to the creation of complex, yet easy to encode models that effectively combine uncertainty and logic. Learning the structure of these models from data reduces the human effort of identifying the right structures. In this work, we present an asynchronous deep reinforcement learning algorithm to automatically learn HL-MRF clause structures. Our algorithm possesses the ability to learn semantically meaningful structures that appeal to human intuition and understanding, while simultaneously being able to learn structures from data, thus learning structures that have both the desirable qualities of interpretability and good prediction performance. The asynchronous nature of our algorithm further provides the ability to learn diverse structures via exploration, while remaining scalable. We demonstrate the ability of the models to learn semantically meaningful structures that also achieve better prediction performance when compared with a greedy search algorithm, a path-based algorithm, and manually defined clauses on two computational social science applications: i) modeling recovery in alcohol use disorder, and ii) detecting bullying.

In today's busy modern life, modeling and accurately predicting fitness center equipment usage and availability is essential for improving human fitness and well-being as it provides people the flexibility to plan their schedule and exercise at their convenience. In addition to its crucial role in ensuring a healthy and sustainable future, adopting a data-driven approach for modeling and predicting fitness center equipment usage is necessary for planning the optimal square footage for developing a fitness center, and determining the kinds of equipment to purchase and install. In this paper, we develop DeepFit, a deep learning based system that predicts future fitness center equipment usage based on historical data. To this end, we design a Long Short Term Memory (LSTM) based sequence-to-sequence model that captures the dependencies in the data. The sequence-to-sequence model comprises of an encoder and a decoder, each of which separately is a deep Recurrent Neural Network (RNN). The basic cell structure in the RNN architecture is an LSTM cell.

The possibility of anonymity and lack of effective ways to identify inappropriate messages have resulted in a significant amount of online interaction data that attempt to harass, bully, or offend the recipient. In this work, we perform a fine-grained quantitative and qualitative linguistic analysis of messages exchanged using one such recent web/smartphone application-Sarahah, that allows friends to exchange messages anonymously. We first develop a weakly supervised hierarchical framework using seeded topic models to automatically categorize Sarahah messages into different coarse and fine-grained bullying categories. Our linguistic analysis reveals that a significant number of messages exchanged using this platform (~20%) include inappropriate, hurtful, or profane language intended to embarrass, offend, or bully the recipient. We then present a detailed analysis of the messages and corresponding users' responses to these messages in the different bullying categories by comparing them across different linguistic and psychological attributes such as sentiment and psycho-linguistic categories from Linguistic Inquiry Word Count (LIWC). Finally, we perform a comparative analysis of messages exchanged on Sarahah to an existing labeled cyberbullying dataset from the Formspring social network on the severity of bullying, coarse-grained bullying categories, and anonymity. Our analysis sheds light on the different categories of bullying and the effect each category has on the recipient and helps quantify the different types and amounts of negativity existing in online social media.

Cities have limited resources that must be used efficiently to maintain their smooth operation. To facilitate efficient resource allocation and management in cities, in this paper, we study one such important problem: how long does it take to resolve non-emergency 311 service requests? We present NYCER, a Non-emergency Response prediction system based on a recently developed structured regression model, sparse Gaussian conditional random fields (GCRFs), that successfully captures the dependencies between historical and future response times. Through extensive experimentation on 311 service requests in New York City (NYC) over a three and a half year period between Jan 2015 to June 2018, we demonstrate that our trained system is able to accurately predict future response times one week in advance using just the previous two weeks data at test time. NYCER achieves superior prediction performance across all agencies, complaint types, and locations, when compared to a linear regression baseline (up to a factor of 2X). The trained NYCER system requires low computational resources and data at test time, thus making it an attractive system that can be readily deployed in practice.

Water availability and management is an important problem plaguing many developing and under-developed countries. Many factors including geographic, political, management, and environmental factors affect the availability of water in these regions. In this paper, we develop an ensemble-learning based predictive-analytics framework for smart water management to predict: i) water pump operation status (e.g., functional, non functional), ii) water quality, and iii) quantity. In the predictive-analytics framework, we first perform feature engineering to select relevant features, use them to develop the XGBoost and Random Forest ensemble learning models, and then perform extensive feature analysis to identify the most predictive features, for each prediction problem mentioned above. We evaluate our framework on two publicly available smart water management datasets pertaining to Tanzania and Nigeria and show that our proposed models outperform several baseline approaches, including logistic regression, SVMs, and multi-layer perceptrons in terms of precision, recall and F1 score. We also demonstrate that our models are able to achieve a superior prediction performance for predicting water pump operation status for different water extraction methods. We conduct a detailed feature analysis to investigate the importance of the various feature groups (e.g., geographic, management) on the performance of the models for predicting water pump operation status, water quality and quantity. We then perform a fine-grained feature analysis to identify how individual features, not just feature groups, impact performance. We identify that among individual features, location (x, y, z coordinates) has the maximum impact on performance. Our analysis is helpful in understanding the types of data that should be collected in future for accurately predicting the different water problems.

Reducing the carbon footprint of energy generation is an important part of ongoing sustainability efforts. To cut carbon footprints, electric utilities are incentivizing renewable energy integration through net metering and introducing time-of-use pricing plans to cut demand peaks, as peaks significantly contribute to both generation costs and carbon emissions. Net metering is one of the most popular means of integrating distributed renewable generation in the grid. However, the current net metering approach doesn't effectively cut demand peaks because renewable harvest peak and demand peaks are out of sync. Furthermore, as several states impose net metering subscriber limits of less than 1% of the peak, net metering isn't even close to realizing the full potential of renewable integration in the grid. To address these limitations, we present GreenPeaks, an energy storage based renewable integration system to enhance net metering. GreenPeaks employs energy storage to intelligently move a fraction of harvested energy to peak intervals and accumulate any surplus harvest. We evaluate GreenPeaks using consumption data from real homes. Our results show that GreenPeaks reduces grid-wide peak by 12% in contrast to net metering's 2%, while reducing the electricity generation costs by more than 40%.

Alcoholism, also known as Alcohol Use Disorder (AUD), is a serious problem affecting millions of people worldwide. Recovery from AUD is known to be challenging and often leads to relapse at various points after enrolling in a rehabilitation program such as Alcoholics Anonymous (AA). In this work, we take a structured approach to understand recovery and relapse from AUD using social media data. To do so, we combine linguistic and psychological attributes of users with relational features that capture useful structure in the user interaction network. We evaluate our models on AA-attending users extracted from the Twitter social network and predict recovery at two different points---90 days and 1 year after the user joins AA, respectively. Our experiments reveal that our structured approach is helpful in predicting recovery in these users. We perform extensive quantitative analysis of different groups of features and dependencies among them. Our analysis sheds light on the role of each feature group and how they combine to predict recovery and relapse. Finally, we present a qualitative analysis of the different reasons behind users relapsing to AUD. Our models and analysis are helpful in making meaningful predictions in scenarios where only a subset of features are available and can potentially be helpful in identifying and preventing relapse early.

Professional networks are a specialized class of social networks that are particularly aimed at forming and strengthening professional connections and have become a vital component of professional success and growth. In this paper, we present a holistic model to jointly represent different heterogenous relationships between pairs of individuals, user actions and their respective propagations to characterize influence in online professional networks. Previous work on influence in social networks typically only consider a single action type in characterizing influence. Our model is capable of representing and combining different kinds of information users assimilate in the network and compute pairwise values of influence taking the different types of actions into account. We evaluate our models on data from the largest professional network, LinkedIn and show the effectiveness of the inferred influence scores in predicting user actions. We further demonstrate that modeling different user actions, node features, and edge relationships between users leads to around 20% increase in precision at top k in predicting user actions, when compared to the current state-of-the-art model.

Arti Ramesh, Shachi H. Kumar, James Foulds, Lise Getoor. Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015.

Maintaining and cultivating student engagement is critical for learning. Understanding factors affecting student engagement will help in designing better courses and improving student retention. The large number of participants in massive open online courses (MOOCs) and data collected from their interaction with the MOOC open up avenues for studying student engagement at scale. In this work, we develop a framework for modeling and understanding student engagement in online courses based on student behavioral cues. Our first contribution is the abstraction of student engagement types using latent representations and using that in a probabilistic model to connect student behavior with course completion. We demonstrate that the latent formulation for engagement helps in predicting student survival across three MOOCs. Next, in order to initiate better instructor interventions, we need to be able to predict student survival early in the course. We demonstrate that we can predict student survival early in the course reliably using the latent model. Finally, we perform a closer quantitative analysis of user interaction with the MOOC and identify student activities that are good indicators for survival at different points in the course.

Maintaining and cultivating student engagement is a prerequisite for MOOCs to have broad educational impact. Understanding student engagement as a course progresses helps characterize student learning patterns and can aid in minimizing dropout rates, initiating instructor intervention. In this paper, we construct a probabilistic model connecting student behavior and class performance, formulating student engagement types as latent variables. We show that our model identifies course success indicators that can be used by instructors to initiate interventions and assist students.