Search

Speech emotion recognition (SER) is one of the most important and active areas of. research in speech processing. Numerous approaches have been proposed to address various limitations in this field, but the sheer diversity of speech emotions, as well as their complexity, continue to make SER a tough nut to crack. This paper attempts to conduct a thorough investigation into speech emotion recognition in order to determine the most appropriate feature set and model for SER. A multi-layer perceptron (MLP) and convolutional neural network (CNN) based ensemble model for SER is proposed, which is a simple yet very powerful model for SER that can greatly improve classification accuracy. The model's performance is evaluated based on four benchmark datasets, namely RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song), EmoDB (Emotional Dat0abase), SAVEE (Surrey Audio-Visual Expressed Emotion), and TESS (Toronto Emotional Speech Set). The proposed model dominates over several baseline methods (decision tree (DT), random forest (RF), support vector machine (SVM), k-nearest neighbour (KNN), and the base learners, i.e., MLP and CNN) in terms of various performance metrics for all the datasets. Furthermore, the proposed model outperforms all previous works for RAVDESS (Acc=73.1%), SAVEE (Acc=83.8%), and TESS (Acc=99.9%) datasets. [ABSTRACT FROM AUTHOR]

Road accident detection and prevention is one of the most challenging problems in the research field revolving around a multitude of problems that need to be addressed. In this context, the research includes creating an automated system to extract geosocial data from multiple sources like news articles and leverage the power of deep learning techniques to classify the collected data as road accidents and extract the useful metadata such as time and date of road accident, location of accident, etc. The news articles are classified into three classes namely non-road accident news, particular road accident, and generic road accident using Word2Vec-based and LSTM-based models. For the combination of LSTM model with K(5)-Fold CV, the road accident news classification model obtains the highest accuracy of about 96%. For location metadata extraction, BERT (Bidirectional Encoder Representations from Transformers)-based baseline models (DBBCS, DTRS, DRBS, DRLS, BLUWwmFS, BLCWwmFS, DBUDS, DBCDS) have also been compared based on various performance metrics such as exact match (EM) and F1-scores. The BLCWwmFS (bert-large-cased-whole-word-masking-finetuned-squad) is found to be the best performing model with an EM and F1-scores of 0.645 and 0.673 respectively for location metadata extraction. [ABSTRACT FROM AUTHOR]

Influence maximization is a task in social network analysis that involves selecting a group of k individuals, known as the "seed set," from the network to maximize the projected number of users influenced, termed as "influence spread." The most important aspect of the influence maximization problem is identifying the most impactful communities and nodes within the network, rather than selecting notable nodes from the entire network. Indeed, unlike most existing approaches, influential node selection as a full network does not assure influence maximization in all clusters. This paper selects seed nodes with elevated betweenness centrality and eigencentrality within each community, communities are identified through the Girvan–Newman method, for influence propagation. The study employs the linear threshold and independent cascade models to assess the speed of influence propagation. Results suggest that choosing seed nodes from each community using these centrality measures is more effective than randomly selecting nodes from the entire network and also from the communities. Moreover, this study is beneficial even if nodes are selected using centrality analysis from the entire network. This approach can help to manage the spread of influence and improve influence maximization in community-structured social networks. [ABSTRACT FROM AUTHOR]

Music has always been the most powerful medium to express human emotion and feeling which sometimes mere words cannot express. As a result generating music using machine and deep learning approaches have been quite popular for some time now. It is a very challenging and interesting task to do as imitating human creativity is not easy. This paper attempted to perform effective melody generation using sequential deep learning models particularly LSTMs (Long short-term memory). In this context, note that the previous works exhibit two principal limitations. Firstly, a significant majority of the studies rely on RNN variants that cannot effectively remember long past sequences. Secondly, they often don't consider the varying temporal context lengths in melody generation during data modeling. In this work, experiments have been performed with different LSTM variants namely Vanilla LSTM, Multi-Layer LSTM, Bidirectional LSTM and different temporal context lengths for each of them to find out the optimal LSTM model and the optimal timestep for efficient melody generation. Moreover, ensembles of the best-performing techniques for each genre (e.g., classical, country, jazz, and pop) are implemented to see if we can generate even better melodies than the corresponding individual models. Finally, a qualitative evaluation is carried out for the generated melodies by conducting a survey that we circulated among fellow colleagues and within the ISMIR community and asked the participants to rate each audio on a scale of 1-5 which helped us in assessing the quality of our generated music samples. All the models have been validated on four datasets that we manually prepared based on their genres namely Classical, Jazz, Country and Pop. [ABSTRACT FROM AUTHOR]

The missing readings in various sensors of air pollution monitoring stations is a common issue. Those missing sensor readings may greatly influence the performance of monitoring and analysis of air pollution data. To address this problem, in this paper, a multi-view based missing value (MV) imputation method called MVDI (Multi-View Data Imputation) is proposed for air pollution related time series data. MVDI combines four models namely LSTM (Long-Short Term Memory), IDS (Inverse Distance Squared), SVR (Support Vector Regressor), and KNN (K-Nearest Neighbors) to estimate MVs. These four models are mainly employed to capture the variations in data from different views of the dataset. Here, different views represent different portions (subsets) of the actual dataset. The estimates of MVs from all the views are combined using a kernel function to get an overall result. The proposed model MVDI is evaluated on real-world air pollution dataset in terms of RMSE, MAE, MAPE, and R2. The experimental results show that MVDI dominates over the baseline methods namely AR (AutoRegressive), ARIMA (AutoRegressive Integrated Moving Average), RFR (Random Forest Regressor), ANN (Artificial Neural Network), LI (Linear Interpolation), NN (Nearest Neighbors), MI (Mean Imputation), CNN (Convolutional Neural Network), ConvLSTM (Convolutional LSTM). [ABSTRACT FROM AUTHOR]

Research Data Management (RDM) has gained significant traction in recent years, being essential to allowing research data to be, e.g., findable, accessible, interoperable, and reproducible (FAIR), thereby fostering collaboration or accelerating scientific findings. We present solutions for RDM developed within the DFG-Funded Cluster of Excellence EXC2075 Data-Integrated Simulation Science (SimTech). After an introduction to the scientific context and challenges faced by simulation scientists, we outline the general data management infrastructure and present tools that address these challenges. Exemplary domain applications demonstrate the use and benefits of the proposed data management software solutions. These are complemented by additional measures for enablement and dissemination to foster the adoption of these techniques. [ABSTRACT FROM AUTHOR]

The sensors available in the smartphones are useful to explore a diverse range of city dynamics (e.g. noise pollution, road condition, traffic condition, etc.). The potential of the smartphone sensors coupled with their widespread availability help to emerge a new paradigm of sensing known as participatory sensing. It uses the power of smartphone equipped sensors to collect, store, and analyze data with high spatiotemporal granularity. In a participatory sensing based system, a task provider (also known as a crowdsourcer) may have a set of sensing tasks regarding different dynamics of a city. Here, adequate users' participation is necessary to acquire a sufficient amount of data which is a key factor for the participatory sensing based systems to provide good service quality. The task providers appoint a set of task executors (smartphone users i.e. participants of crowdsensing tasks) to execute those sensing tasks. But, existing works on sensing task allocation suffer from lack of good incentive mechanisms that are attractive for the task executors. In order to address this issue, in this paper, a double auction based incentive mechanism called TATA (Truthful Double Auction for Task Allocation) is proposed for participatory sensing. TATA performs fair allocation of tasks which is leading to efficient incentive mechanism. In the case of TATA, the fair allocation of sensing tasks of the task providers to the task executers indicates that the proposed double auction mechanism is able to satisfy the truthfulness property in order to resist market manipulation (i.e., untruthful bidding and asking). Specifically, TATA achieves all the desirable properties like individual rationality, truthfulness (i.e. incentive compatibility), budget balance, etc. TATA is also computationally efficient and yields high system efficiency. Additionally, the performance of the proposed incentive mechanism is evaluated and compared with the existing mechanisms through extensive simulations based on the real-world data from Amazon Mechanical Turk. TATA yields high utility and satisfaction for the task providers and executors as compared to the existing mechanisms. [ABSTRACT FROM AUTHOR]

The honking events' density reflects the level of traffic noise pollution, road congestion, etc in the urban areas. In this paper, we propose a participatory sensing based traffic honk monitoring system called HonkSense that uses smartphone equipped sensors (e.g. microphone, GPS, etc.). Citizens can take part in monitoring traffic noise pollution due to honking by recording ambient noise on the road. Application running on users' smartphones is used to extract features in real time from recorded audio and then send to the cloud for honk detection and decision making tasks. Here, Mel-Frequency Cepstral Coefficients (MFCCs) are utilized as feature for presenting audio signals in honk detection. This paper uses a deep Convolutional Neural Network (CNN) model that is deployed to cloud for detecting traffic honking events. The end-to-end system provides a privacy-preserving (anonymous data collection), low-power and low-cost solution for participatory sensing based traffic honk monitoring. We evaluate our proposed system on real world participatory sensing based road sound dataset collected by participants. It achieves a classification accuracy of 96.3%. The deep CNN is also evaluated on different benchmark datasets (namely ESC-50 and UrbanSound8K). The results are also compared with the baseline support vector machine (SVM) and k-nearest neighbors (KNN) classification models. Besides, state-of-the-art visualization techniques are used to explore spatial and temporal variability of honking events in urban areas using two case studies. [ABSTRACT FROM AUTHOR]

Road surfaces can contain a lot of anomalies such as potholes. Understanding and detecting such irregularities are a key towards building a safe road transport system and a step towards building a smart city. This paper makes an attempt to classify road surface images using transfer learning on pre-trained convolution neural network (CNN) based models. In this work, two types of transfer learning approaches namely fine tuning and feature extraction are employed on various state-of-the-art CNN architectures of recent past such as VGG-16 (by Visual Geometry Group), VGG-19, Inception, ResNet-50 (Residual Network), Xception, DenseNet-121, SEResNet-50 (Squeeze-and-Excitation ResNet model), and EfficientNet-B3. The performance of the models are evaluated based on real-world dataset of road surface images collected by the participants. The evaluation metrics considered are test accuracy, precision, recall, F1-score, specificity, area under the receiver operating characteristic curve (ROC-AUC) and also training and inference time measurements. It is observed that among all the deep CNN models, SEResNet-50 (F1-score = 0.954, ROC-AUC = 0.993), Xception (F1-score = 0.943, ROC-AUC = 0.987) and DenseNet-121 (F1-score = 0.941, ROC-AUC = 0.987) achieve the best performance in terms of the evaluation metrics while fine tuning is considered with data augmentation. Effectiveness of the deep CNN models are also evaluated on various benchmark datasets, namely Road Traversing Knowledge (RTK) dataset and pothole dataset. [ABSTRACT FROM AUTHOR]

Human activity recognition by the use of smartphone-equipped sensors has gotten a lot of interest in current times because of its large variety of applications.In this regard, this study provides a comprehensive comparative analysis of shallow and deep learning models for smartphone-based HARover high granular daily human activities. Moreover, A robust architecture for smartphone-based HAR is also provided, with stages ranging from data collection to data modelling. A total of seven best performing HAR models namely Decision Tree (DT), Random Forest(RF), DeepNeural Networks (DNN), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Gradient Boosting (GB) and Convolutional Neural Networks (CNN) are investigated. This research work is based on a real-world dataset of 95690 data samples collected from the smartphone sensors of 18 different subjects. The comparative study reveals that three models namely DNN, RF, and GB mostly dominated over the other models in terms of five performance metrics namely accuracy, recall, precision, F1-score, and AUC value. [ABSTRACT FROM AUTHOR]

Air pollution has become a major environmental risk of the new civilized world due to its severe influence on public health and the environment. Eventually, understanding the spatiotemporal variability of air pollution at high granularity is necessary to make relevant public policies. To explore spatiotemporal variability of air pollution at high granularity we have utilized the power of IoT based participatory sensing and data science. In this paper, we propose a predictive model for spatiotemporal air pollution estimation technique called Multiview data Fusion model (MVDF) that can consider spatial as well as temporal dependencies of air pollutants. The proposed technique is evaluated based on real-world air pollution dataset collected by participants over a period of 1 year in an urban area of city Kolkata. The results show that MVDF dominates over some baselines like Simple Kriging (SK), Modified Shepard's Method (MSM) and Nearest Neighbor (NN). Besides, in this paper, we attempt to perform visual analysis that consists of state-of-the-art visualization techniques to explore spatiotemporal variability at different granularities on the estimated pollution levels of MVDF. [ABSTRACT FROM AUTHOR]

Analysis and prediction on real time air quality data is a critical step in solving various problems related to pollution and finding a genuine solution. However, missing values in air pollution data is a serious issue that may greatly influence the performance of such analysis and prediction. In order to address this problem, a 2-step process is proposed, consisting of a data pre-processing stage using Q-FUZZY (Quantized Fuzzification) model and data imputation stage via a Fuzzy Imputation Model(FIM) to handle this complexity. We have validated the proposed approach using a real world dataset of Kolkata, India containing pollution levels of different air pollutants along with meteorological parameters. Various performance measures are used to determine the effectiveness of FIM. Moreover, the performance of the proposed model is also compared with the baselines namely, FLAR (Fuzzy Least Absolute Regression), ridge regression and EMD-SVR-SARIMA (Empirical Mode Decomposition-Support Vector Regressions-Seasonal Autoregressive Integrated Moving Average). On comparison, it is observed that FIM achieves an overall better performance in terms of distance measures, Mean Similarity Measures (MSM), Mean Inclusion Measures (MIM) and Mean Predictive Ability (MPA). [ABSTRACT FROM AUTHOR]

Existence of several challenges and high cost in the development of monitoring infrastructure have become major reasons for data sparsity by statutory government agencies tasked to study pollution exposure in urban areas. As an effort to mitigate this problem, the recent usage of satellite aerosol optical depth data along with the usage of learning algorithms have become popular in recent times. This paper presents a novel four-staged approach using different machine learning, deep learning and statistical methods to develop a spatio-temporal hybrid model for temporal forecasting using data from existing stations along with satellite aerosol optical depth data for spatial interpolation. Experiments conducted on real-world data belonging to the cities of Kolkata, Bengaluru and Mumbai show that a consistent pattern is not followed in all the cities in all stages except in spatial interpolation where Random Forest Regression is found to surpass all other models used. While a long short-term memory network (LSTM Auto-Encoder) when employed in temporal forecasting inside the hybrid method outperforms others in Mumbai, a random forest regression-based method and a multi-layer perceptron-based method outperform others similarly in Kolkata and Bengaluru, respectively. [ABSTRACT FROM AUTHOR]

Emotion detection is a promising field of research in multiple perspectives such as psychology, marketing, network analysis and so on. Multiple models have been suggested over the years for accurate and efficient mood detection. Identifying emotion, or mood, from text has progressed from a simple frequency distribution analysis to far more complicated learning approaches. The main aim of all these text mining and analysis is twofold. First is to categorise existing text into broad classes of emotions, such as happy, sad, angry, surprised and so on. The second aim is to accurately predict the moods of real-time streaming text. The novelty of the work lies in the extensive comparison of nine conventional learning methods with respect to performance metrics precision, recall, F1 and accuracy as well as studying the variance of mood over time using a wide array of moods (25). Using conventional classifiers allow near real-time predictions, can work on considerably less training data, and has the flexibility of feature engineering, as deep learning methods have feature engineering embedded in the model. Since a single line of text can be associated with multiple emotions, this article compares the performance of classifiers in predicting multiple moods for streaming text with likelihood-based ranking. An android application named Citizens' Sense was developed for text collection and analysis. The performance of mood classifiers are tested further using Twitter data related to COVID19. Based on the precision, recall, F1 and accuracy of the classifiers, it can be seen that Random Forest, Decision Tree and Complement Naive Bayes classifiers are marginally better than the other classifiers. The variance of mood over time, and predicted moods for text support this finding. [ABSTRACT FROM AUTHOR]

One of the most challenging tasks in time-series prediction is a model's capability to accurately learn the repeating granular trends in the data's structure to generate effective predictions. Traditionally specially tuned statistical models and deep learning models like recurrent neural networks and long short-term memory networks are used to tackle such problem of sequence modeling. However in practice, factors like inadequate parameters in case of statistical models, random weight initializations, and data inadequacy in case of deep learning models affect the resulting final predictions. As a possible solution to these known problems, this paper introduces a novel method of time-series labeling (TSL) comprising a combination of encoding and decoding methodologies that not only takes into account the granular structure of a time-series data but also its underlying meta-learners for better predictive accuracy. To demonstrate the approach's effectiveness and capability of handling wide range of scenarios, comparisons are drawn first over different widely used statistical and deep learning models and then applying TSL to each of them in order to showcase the resulting performance improvement when implemented over a wide variety of real-world datasets. The experimental findings reflect an average of 25% increase in overall performance when using TSL along with mostly similar performance of different combinations regardless of model complexity thereby proving its efficacy in predicting periodic data. [ABSTRACT FROM AUTHOR]

The behaviour of individual users in an online social network is a major contributing factor in determining the outcome of multiple network phenomenon. Group formation, growth of the network, information propagation, and rumour blocking are some of the many network behavioural traits that are influenced by the interaction patterns of the users in the network. Network motifs capture one such interaction pattern between users in online social networks (OSNs). For this work, four second-order (two-edged) network motifs have been considered, namely, message receiving pattern, message broadcasting pattern, message passing pattern, and reciprocal message pattern, to analyse user behaviour in online social networks. This work provides and utilizes a node interaction pattern-finding algorithm to identify the frequency of aforementioned second-order network motifs in six real-life online social networks (Facebook, GPlus, GNU, Twitter, Enron Email, and Wiki-vote). The frequency of network motifs participated in by a node is considered for the relative ranking of all nodes in the online social networks. The highest-rated nodes are considered seeds for information propagation. The performance of using network motifs for ranking nodes as seeds for information propagation is validated using statistical metrics Z-score, concentration, and significance profile and compared with baseline ranking methods in-degree centrality, out-degree centrality, closeness centrality, and PageRank. The comparative study shows the performance of centrality measures to be similar or better than second-order network motifs as seed nodes in information diffusion. The experimental results on finding frequencies and importance of different interaction patterns provide insights on the significance and representation of each such interaction pattern and how it varies from network to network. [ABSTRACT FROM AUTHOR]

Forecasting time series has acquired immense research importance and has vast applications in the area of air pollution monitoring. This work attempts to investigate the abilities of various existing techniques when applied for short term, high granular time series forecasting of PM2.5. More specifically, a comparative study has been provided, taking into account both popularly used models and lesser-used models in this area. The study has been carried out considering ten well defined models that are ARIMA (auto-regressive integrated moving average), SARIMA (seasonal ARIMA), SES (single exponential smoothing), DES (double exponential smoothing), TES (triple exponential smoothing), ANN (artificial neural network), DT (decision tree), kNN (k-nearest neighbor), LSTM (long short-term memory) and MCFO (markov chain first order). A framework has been built that categories the models, implements them under identical execution environment and forecasts succeeding values. Implementation has been carried out over five data sets of real-world air pollution time series, that are collected from five differently located government setup monitoring stations over a period of 1 year (July 2018-June 2019). Rigorous statistical analysis has been performed that yields an insight to the nature and variability of these time series data. Forecasting has been carried out on short term basis, focusing on high granularity whereas, three different lengths of forecast horizon (1 day, 1 week, and 1 month) have been tested. Eventually, the models have been compared in terms of their associated performance measuring units namely, RMSE (root mean of squared error), MAE (mean absolute error) and MAPE (mean absolute percentage error). The comparative results verified with multiple datasets show that all the models posses less error for a shorter forecast horizon, where LSTM providing the best performance. Superiority of machine learning and deep learning models are found in case of longer length of forecast horizon with kNN achieving best accuracy whereas, significant performance degradation of ARIMA is found for longer forecast horizon. Moreover, TES, DT, kNN, LSTM, MCFO are found to be well adopted in relation with shape and variability of the data. Note that the performance on various length of high granular forecast horizon have been studied over multiple datasets that give an added value to this work. [ABSTRACT FROM AUTHOR]

The growing market of cloud computing resulted in increased demand for cloud resources and it will become difficult for individual service providers (SPs) to fulfill all resource requests. That leads to a situation where two or more SPs may form a group (federation) and share the resources in order to fulfill the cloud users’ demand and gain economic advantage. Now, due to the formation of more than one federations by different cloud providers, it may be difficult for users to select a suitable federation who can deliver cloud services at a fair price. In this context, it is necessary to have a framework that will efficiently allocate resources of cloud federations to the users at a fair price and stop market manipulation. In this article, we propose a multi-unit double auction mechanism called TARA (Truthful Double Auction for Resource Allocation) that can be used to efficiently choose cloud federations for users from which they can get resources. Here, we consider a multi-seller and multi-buyer double auction mechanism for heterogeneous resources, where every buyer submits their bids and every seller places their ask (the price of a resource that is offered by a federation). TARA achieves some important properties like truthfulness (also known as incentive compatibility), individual rationality and budget balance for both buyers and sellers. TARA is also computationally efficient and posses high system efficiency. The simulation results also show that total utility of buyer is more than some existing double auction mechanisms. [ABSTRACT FROM AUTHOR]

In smartphone-based crowd/participatory sensing systems, it is necessary to identify the actual sensor data provider. In this context, this paper attempts to recognize the users' identity based on their gait patterns (i.e. unique walking patterns). More specifically, a deep convolution neural network (CNN) model is proposed for the user identification with accelerometer data generated from users smartphone sensors. The proposed model is evaluated based on the real-world benchmark dataset (accelerometer biometric competition data) having a total of 387 users accelerometer sensor readings (60 million data samples). The performance of the proposed CNN-based approach is also compared with five baseline methods namely Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Logistic Regression (LR), and K-Nearest Neighbours (KNN). It is observed that the proposed model achieves better results (accuracy = 98.8%, precision = 0.94, recall = 0.97, and F1-score = 0.95) as compared to the baseline methods. [ABSTRACT FROM AUTHOR]

As the number of enterprises dispatching their workload to the cloud has increased significantly over the last decade, service level agreements (SLAs) becoming a key element to consider for maintaining the quality of service (QoS). In order to facilitate the perseverance of service quality at a satisfactory level, clouds perform load balancing through migration of virtual machines (VMs) from overloaded physical machines (PMs). However, there are several challenges in achieving effective and efficient load balancing. First, VMs in clouds use different resources to serve a variety of applications, which results in varying levels of resource overutilization in different PMs. Second, due to the application’s time-varying heterogeneous nature of resource requirements, the PM’s resource consumption vary over time, making the profiling of resources difficult. Migration decisions in previous load balancing techniques are mostly based on deterministic resource demand estimation, which treats each resource equally and leads to inefficient migrations, causing severe SLA violations in terms of performance degradation. To address this problem, we propose a SLA-aware stochastic load balancing scheme using VM migrations, namely SLA-LB. It provides probabilistic guarantee against resource overloading, while satisfying the SLA. As opposed to previous methods, SLA-LB dynamically assigns different weights to different resources based on PM’s overload probability and effectively addresses the multidimensional resource requirement with stochastic characterization. Experimental results using PlanetLab and Google Cluster trace show that SLA-LB outperforms previous load balancing methods, i.e., RIAL, Sandpiper and CloudScale in terms of performance degradation by an average margin of 10.8%, 23.53% and 33%, respectively. [ABSTRACT FROM AUTHOR]

Tackling air pollution has become of utmost importance since the last few decades. Different statistical as well as deep learning methods have been proposed till now, but seldom those have been used to forecast future long-term pollution trends. Forecasting long-term pollution trends into the future is highly important for government bodies around the globe as they help in the framing of efficient environmental policies. This paper presents a comparative study of various statistical and deep learning methods to forecast long-term pollution trends for the two most important categories of particulate matter (PM) which are PM2.5 and PM10. The study is based on Kolkata, a major city on the eastern side of India. The historical pollution data collected from government set-up monitoring stations in Kolkata are used to analyse the underlying patterns with the help of various time-series analysis techniques, which is then used to produce a forecast for the next two years using different statistical and deep learning methods. The findings reflect that statistical methods such as auto-regressive (AR), seasonal auto-regressive integrated moving average (SARIMA) and Holt–Winters outperform deep learning methods such as stacked, bi-directional, auto-encoder and convolution long short-term memory networks based on the limited data available. [ABSTRACT FROM AUTHOR]

The demand of cloud-based services is growing rapidly due to the high scalability and cost-effective nature of cloud infrastructure. As a result, the size of the data center is increasing drastically, so is the cost of maintenance in terms of resource management and energy consumption. Hence, it is important to develop a proper resource management plan to maximize the profit by reducing the overhead of operational cost. In this paper, we propose a multi-step-ahead workload prediction approach using Machine learning techniques and allocate the resources based on this prediction in a way that allows the resources to be utilized more efficiently and thereby, reducing the data center's overall energy consumption. We evaluate the effectiveness of our framework based on real workload trace of Bitbrains. Experimental results show that our framework outperforms other state-of-the-art approaches for predicting workload over a long-run and significantly improves resource utilization while enabling substantial energy savings. [ABSTRACT FROM AUTHOR]

Proper maintenance of roads is an extremely complex task and also an important issue all over the world. One of the most critical road monitoring and maintenance activities is the detection of road anomalies such as potholes. Identification of potholes is necessary to avoid road accidents, prevent damage of vehicles, enhance travelling comforts, etc. Although maintenance of roads is considered to be a serious issue by the authorities over the years, lack of proper detection and mapping of road potholes makes the issue more severe. To overcome this problem, an end-to-end system called PotSpot is built for real-time detection, monitoring, and spatial mapping of potholes across the city A Convolutional Neural Network (CNN) model is proposed and evaluated on real-world dataset for pothole detection. Additionally, real-time pothole-marked maps are generated with the help of Google Maps API (Application Programming Interface). To provide an end-to-end service through this system, both the pothole detection and pothole mapping are integrated through an android application. The proposed model is also compared with six baselines namely Artificial Neural Network (ANN), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and three pre-trained CNN models InceptionV3, VGG19 and VGG16 in terms of performance metrics to verify its effectiveness. The proposed model achieves better accuracy (≈ 97.6 %) as compared to the above-mentioned baseline methods. It is also observed that the Area Under the Curve (AUC) value for the proposed pothole detection model (AUC= 0.97) is higher than the baseline methods. [ABSTRACT FROM AUTHOR]

Indoor air pollutants e.g., Carbon dioxide (CO2), Particulate Matter(PM)2.5, PM10, Total Volatile Organic Compounds (TVOC), etc. have a serious impact on human health. Out of these pollutants, CO2 is one of the most dominant one. Hence, proper monitoring and control of this pollutant is an important part of maintaining a healthy indoor. To make this happen, it is required to predict the next moment's indoor pollutant level at an acceptable accuracy range that ensures necessary steps can be taken beforehand to avoid a rise in the indoor pollution level for maintaining a healthy indoor all the time. It also helps people plan ahead, decreases the adverse effects on health and the costs associated. For this experiment, we have collected three months of real-life time-series data along with proper context information and have gone through feature engineering and feature selection process to create model ready data. Now, since the indoor CO2 concentration is dependent on multiple external factors (context data) which in turn is dependent on time, makes it a time-dependent function. Hence, to predict the indoor pollutant CO2, here we have used the time series forecasting model based on our collected data nature. This is a powerful tool and used in a wide range of research domains for predicting the next moment's target value. This model ready data is utilized in forecasting different time series models. According to our findings, among the selected popular time series models, the SARIMAX time series model is best suited for this forecasting problem which is utilizing indoor context information along with historical data (with 10 Fold Time-Series Split Cross-Validation score 0.907). We have achieved an average of RMSE 26.45 ppm (i.e., 97.36% accuracy) level based on a three day average for indoor pollutant prediction which is outperforming other relevant models in this domain. [ABSTRACT FROM AUTHOR]

English proficiency is considered a door opener almost in any field anywhere in the world, yet all the attempts to improve effectiveness of teaching and learning of English language in Bangladesh seem to fail over decades. Although poor learning motivation is counted as a key aspect why students do not hold their attention to their English classes and make use of their capabilities to understand and learn cognitively, there has been a very limited number of studies in Bangladeshi context to deal with this critical issue. This mixed method study attempted to identify the sources of motivation for attending English classes of the 1st year undergraduate students at a Bangladeshi university. Quantitative data was collected using two close-ended questionnaires from 50 randomly selected Bangladeshi tertiary level English language learners and 10 teachers. The student questionnaire was developed by adding three items with the ten items adapted from Clement et al. (1994) and the teachers’ questionnaire was developed and administered to follow up the findings from the student participants. For a more complete analysis, structured interviews were conducted with 10 purposively selected students. The results indicated lack of intrinsic motivation among the learners, and based on the existing literature, the study recommended the way outs to improve motivation of the non-performing learners. The recommendations included, shifting from exam-oriented assessment system, teachers’ reflection of their own teaching practices, redesigning course curricula based on learning need and learners’ expectations, ensuring easy access to the latest educational facilities, and encouraging the roles of students and their peers to develop learners’ motivation and foster their autonomous learning opportunities. Finally, the limitations are drawn and suggestions for further research are provided. [ABSTRACT FROM AUTHOR]

COVID-19 is a global crisis where India is going to be one of the most heavily affected countries. The variability in the distribution of COVID-19-related health outcomes might be related to many underlying variables, including demographic, socioeconomic, or environmental pollution related factors. The global and local models can be utilized to explore such relations. In this study, ordinary least square (global) and geographically weighted regression (local) methods are employed to explore the geographical relationships between COVID-19 deaths and different driving factors. It is also investigated whether geographical heterogeneity exists in the relationships. More specifically, in this paper, the geographical pattern of COVID-19 deaths and its relationships with different potential driving factors in India are investigated and analysed. Here, better knowledge and insights into geographical targeting of intervention against the COVID-19 pandemic can be generated by investigating the heterogeneity of spatial relationships. The results show that the local method (geographically weighted regression) generates better performance ( R 2 = 0.97 ) with smaller Akaike Information Criterion (AICc = - 66.42 ) as compared to the global method (ordinary least square). The GWR method also comes up with lower spatial autocorrelation (Moran's I = - 0.0395 and p < 0.01 ) in the residuals. It is found that more than 86% of local R 2 values are larger than 0.60 and almost 68% of R 2 values are within the range 0.80–0.97. Moreover, some interesting local variations in the relationships are also found. [ABSTRACT FROM AUTHOR]

Energy harvesting facilitates Wireless Sensor Networks (WSN) to work in perpetual mode. But, the amount and duration of green energy depend on the unpredictable behavior of ambient energy sources. Limited knowledge of future energy availability is the main constraint in designing routing and MAC (Medium Access Control) protocol. If the prediction of future energy availability can be done with fine accuracy, then energy management strategies can be designed accordingly. Predicting solar irradiance value is a challenging issue as it has seasonal and daily temperature variation. Enhanced-Pro, a novel energy prediction strategy is proposed here by considering past energy observations as well as current energy intake, and its effectiveness is also tested. It is a modification of Pro-Energy, which is a landmark solar energy prediction solution in terms of accuracy. In this work, real-life solar traces are used for prediction. The algorithm incorporates two factors tuning factor and fine adjustment index. These two factors help to enhance prediction accuracy. Enhanced-Pro is compared with Pro-Energy, I-Pro-Energy, and QL-SEP. The simulation result proves Enhanced-Pro delivers better performance in the scale of computational complexity and accuracy. [ABSTRACT FROM AUTHOR]

Currently, all online social networks (OSNs) are considered to follow a power-law distribution. In this paper, the degree distribution for multiple OSNs has been studied. It is seen that the degree distributions of OSNs differ moderately from a power law. Lognormal distributions are an alternative to power-law distributions and have been used as best fit for many complex networks. It is seen that the degree distributions of OSNs differ massively from a lognormal distribution. Thus, for a better fit, a composite distribution combining power-law and lognormal distribution is suggested. This paper proposes an approach to find the most suitable distribution for a given degree distribution out of the six possible combinations of power law and lognormal, namely power law, lognormal, power law–lognormal, lognormal–power law, double power law, and double power law lognormal. The errors in the fitted composite distribution and the original degree distribution of the OSNs are observed. It is seen that a composite distribution fitted using the approach described in this paper is always a better fit than both power-law and lognormal distributions. [ABSTRACT FROM AUTHOR]

Participatory sensing has become an effective way of sensing urban dynamics due to the widespread availability of smartphones among citizens. Traditionally, separate urban sensing applications are designed to monitor different urban dynamics like environment, transportation, mobility, etc. However, combining these applications to aggregate information can lead to various new inferences. The main objective of this work is to improve urban sensing applications by overcoming their individual limitations. A unified framework called JUSense (Judicious Urban Sensing) is proposed that can derive benefits from these applications by combining their functionalities. JUSense provides the opportunity for applications to tackle the challenges associated with data collection, aggregation of data in cloud, calibration, data cleaning, and prediction. A multi-view fusion model is proposed for spatiotemporal urban air and noise pollution map generation. Further, a random forest classifier is built to classify the driving events. Here, large scale experiments are performed to evaluate the efficacy of JUSense on real-world dataset. Both the fusion model and the random forest classifier yield better accuracies compared to the baseline methods. Additionally, case studies are conducted to show the advantages that can arise out of the mutual interactions among the applications. [ABSTRACT FROM AUTHOR]

The technology, cloud computing, in present days, is vastly used due to the services it provides and the ease with which they can be availed. The enormous development of the Internet technology is due to the advent of the concept of cloud. Along with its benefits, cloud computing brings along itself a detrimental side effect, i.e., carbon emission. This is due to the massive energy consumption in the cloud data centers. Reduction in energy consumption in cloud is thus one of the major challenges among the researchers. This work conducts a thorough study of the various techniques that help in minimization of energy consumption in data centers. It also explores and proposes approaches to reduce the same, eventually making the environment greener. In the proposed work, prediction mechanism has been adopted and implemented on the existing Minimization of Migration (MM) policy for large history data set, followed by dynamic thresholding mechanism in place of static thresholds. Rigorous simulations have been conducted, and the results show reduction in cloud data center energy consumption. [ABSTRACT FROM AUTHOR]

In mobile crowd-sensing, smartphone users take part in sensing and then share the data to the server (cloud) and get an incentive. These data can be utilized for providing better services to improve quality of life. Batteries used in smartphones constrain the usability of these devices for longer charge cycles. Hence, maintaining a balance between energy consumption due to crowd-sensing application and that due to the current computational load on the device is the need of the hour. Consequently, in this paper, we formulate strategies applying Markov Decision Process (MDP) by which a smart handheld would crowd-sense while keeping the device active for a longer period of time. MDP used here helps to decide when a device would lend itself to crowd-sense considering the remaining energy of the device, it's recharging probability, current computational load,and the incentive it receives. In this work, we have considered indoor localization as an example of a smartphone based crowd sensing application. The strategies found by solving MDP formulation are implemented for a smartphone application for crowd-sensed indoor localization. We have experimented using 5 smart handheld devices for different use cases. Our scheme is found to perform better than the state-of-the-art works. [ABSTRACT FROM AUTHOR]

Changes in the BAN topology are caused due to the body movement induced by repeated activities like walking, running, twisting, turning and waving arms. In such activities, individual nodes may move relative to each other and along with this, the entire BAN may move its absolute location, which can induce several complexities in the network like re-transmission, collision, overhearing and changing interference. This paper presents a scheme referred as, posture based data transmission (PBDT) with the objective of efficient data transmission. PBDT is based on the occurrence of potential (best) posture over time in repeated activities. In PBDT, each node follows the procedures: (a) recognized the sequence of postures by observing the variation of received signal strength indicator (RSSI) from neighbor nodes over time, (b) finds the best posture from posture sequence for data transmission, (c) maintains a dynamic active/sleep schedule in order to reduce lossy transmission, collision and overhearing. Here, we consider walking as a repeated action over time to check the validity of the proposed mechanism. PBDT is implemented using Castalia simulator and compared with selected MAC protocols. The results are analyzed in respect to the energy consumption and throughput. [ABSTRACT FROM AUTHOR]

Computation offloading effectively expands the usability of mobile terminals beyond their physical limits, and also greatly extends their battery charging intervals. Offloading or cyber foraging is a technique by which large and complex computational jobs are relocated from lightweight portable devices (such as smartphones) called offloadee to powerful servers (such as nearby laptops/desktops or cloud server over the Internet) called surrogates, and getting the output back at the offloadee. Many research works have been done so far on the architecture of offloading systems, but only few works can be found on fault detection and tolerance. So this paper concentrates on categorizing different failures that may affect the benefit of offloading computation and proposes a fault-tolerance approach to alleviate those faults. The checkpoint based fault-tolerance approach proposed in this paper is able to handle crash, omission and transient failures altogether. Fault prevention is obtained by utilizing historical data for choosing worthy surrogate from known neighborhood. This fault tolerance model is evaluated by applying a real application named π calculator and Scimark benchmark suite. The system is implemented and the proposed approach is found to be effective with respect to energy consumption and resource utilization even in case of crash and omission failure. [ABSTRACT FROM AUTHOR]

The emergence of cloud computing has led to an astronomical growth in the computing services provided by vendors over the cloud interface. This has led to the paradigm of cloud federations where a group of CSPs collaborate to form a federation for seamless provisioning of resource requests. In this paper, cloud federation formation framework is modeled as a multi-objective optimization problem with the trade-off between profit and QoS. Federation formation algorithms try to maximize the federation profit while maintaining a balance between the QoS and the profit of the members of the federation. We have applied Linear Scalarization as well as ε-constraint method to find the pareto-optimal solution to this multi-objective optimization problem. A heuristic-based algorithm for cloud federation formation following the integer linear program is proposed. We perform extensive experiments to investigate the performance of our proposed mechanism and show that our proposed mechanism yields optimized solution to the general problem of profit/QoS trade-off. [ABSTRACT FROM AUTHOR]

By embracing the potential of IoT and smartphones, traditional cities can be transformed to smart cities. The success of such smart city mission is firmly vested in populace and thus it should have a bottom-up nature, initiated by the citizens. This paper focuses on the design and development of a unified framework, which can provide a platform to empower all the applications across different dimensions of urban life in a smart city. The aim of this framework is to connect citizens, data, knowledge and services related to IoT as well as smartphone based applications. Here, we categorize all the applications for the smart city in three representative types, viz. IoT based, IoT and smartphone based and smartphone as IoT based applications. We have also developed and tested one prototype following this architecture for each of these three representative category type, i.e, IoT based smart classroom, IoT and smartphone based air quality monitoring system and only smartphone based noise monitoring system to demonstrate the effectiveness of the proposed framework for the smart city scenario. [ABSTRACT FROM AUTHOR]

Home Accidents have been identified as the largest single cause of death after the age of one year and are among the most severe health problems facing the world today. Keeping the toddler safe is a major priority for parents. Kid source online categories about toddler's accidents were home safety from poisoning, electrical appliances, environmental safety from fall and drowning and safety play from injurious toys. Injury prevention at home is an important concern of parents of toddlers. situations that can result in injury In both the industrialized world and developing countries, accidents remain one of the major five leading causes of death. The objective of the study was to assess the level of knowledge regarding prevention of domestic accidents A quantitative approach with descriptive research design was selected for the study. 50 number of mothers were selected by convenience sampling techniques from paediatric OPD. After obtaining the permission from institutional authority the consent from the mothers were taken the knowledge was assessed by using a self administered structured questionnaire. The result showed that greater than 54% of mother are having average knowledge, 30% of mother having poor knowledge and 16% of mother are having good knowledge regarding the prevention of domestic accident. [ABSTRACT FROM AUTHOR]

Cloud federation has paved the way for cloud service providers (CSP) to collaborate with other CSPs to serve users’ resource requests, which are prohibitively high for any single CSP during peak time. Moreover, to entice different CSPs to participate in federation, it is necessary to maximize the profit of all CSPs involved in the federation. Further, federation enables overloaded CSPs to distribute their load among other underloaded member CSPs of federation by migrating the virtual machines (VM). Migration of VM among member CSPs of federation, also enables to increase the reliability and availability of cloud services on occurrence of faults in the datacenters of CSPs. Thus it becomes important for CSPs to form a federation with other CSPs, in such a way that the migration cost of VMs between CSPs of the same federation is minimized and simultaneously profit of CSPs in federation is maximized. In this paper, we model the problem of forming federation among CSPs as a hedonic coalition game, with a utility function depending on profit and migration cost, with the objective of maximizing the former and minimizing the latter. We propose an algorithm to solve this hedonic game and compare its performance with other existing game-theory based cloud federation formation mechanisms. [ABSTRACT FROM AUTHOR]

Measurement-Device-Independent Quantum Private Query (MDI QPQ) with qutrits is presented. We compare the database security and client's privacy in MDI QPQ for qubits with qutrits. For some instances, we observe that qutrit will provide better security for database than qubit. However, when it comes to the question of client's privacy, we have to take additional measures in case of qutrit. Hence, we conclude that though in case of Quantum Key Distribution (QKD) higher dimension provides better security, in case of QPQ this is not obvious. [ABSTRACT FROM AUTHOR]

Wireless sensor network (WSN)-based solutions are highly promising and obvious choice for a wide range of real world monitoring applications. Present studies in WSNs aimed to improve the network characteristics and technologies, and also encouraged timely transformation in WSN. Body area network (BAN) is one such marked change in the form of WSN, which is used in physiological and medical monitoring. Application requirements also induce several challenges in the network, such as node movement, change in propagation medium, and heterogeneous sensing abilities. The presence of these features makes the underlying WSN environment more exigent. We have termed such an environment as mixed WSN (MWSN) in this paper. The main objective of this paper is to highlight the research challenges in BAN due to the presence of MWSN features with a discussion on future direction. [ABSTRACT FROM PUBLISHER]

Energy efficiency is one of the key challenges for designing routing algorithms for wireless sensor networks (WSN). Fast data delivery from source node to sink node is also required for many applications. Sometimes data transmission towards the sink may be interrupted because of failure of nodes in a particular area. The various such requirements of routing data in a WSN are sometimes contradictory and routing algorithms must be designed keeping in mind all these requirements. In this paper three routing algorithms are proposed for multi-sink partitioned network. In these three algorithms, the source nodes or intermediate nodes select a next node to forward the data to the destination or sink. This process repeats until the data reaches the sink. In the first technique, the next node is chosen considering its distance from sink node as a criterion. In the second technique, remaining energy of the neighboring nodes is used as a criterion for selection of the next node. The combination of these two criteria is considered in the third technique and a multiple criteria decision analysis technique is used. A recovery technique from area failure is also proposed and incorporated in the above mentioned algorithms. The algorithms are designed and implemented in multi-sink environment for partitioned wireless sensor networks. Performances of the algorithms are studied in a simulation environment and compared with other well-known algorithms to understand their effectiveness. [ABSTRACT FROM AUTHOR]