Time series forecasting of infectious diseases is crucial to addressing significant global health challenges. The application of AI, particularly Deep Learning (DL) algorithms, has demonstrated substantial success in sequence modeling; however, their performance in epidemiological forecasting remains constrained by the non-stationary nature and complex frequency dynamics of disease transmission. This research introduces a novel Modulated Transformer (FMT) framework to address these challenges. The FMT framework decomposes time series data into distinct frequency-modulated signals, utilizing self-attention mechanisms to capture temporal frequencies. A transformer encoder–decoder architecture then predicts and captures multi-scale temporal dependencies and makes accurate predictions. The novelty of the proposed approach lies in decomposing the input time series into Intrinsic Mode Functions (IMFs) and integrating the frequency-specific components into a Transformer architecture via entropy-based feature selection. The FMT framework significantly reduces Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) by approximately 50% and 65%, respectively, compared to conventional methods. Additionally, it achieves an 8% increase in the R2 score, demonstrating enhanced predictive accuracy. The proposed methodology is evaluated using COVID-19 datasets from multiple countries along with an influenza dataset and benchmarked against statistical, machine learning, and state-of-the-art DL baselines. The contribution of entropy-based IMF integration is systematically examined by comparing results with and without this component, underscoring its importance in improving predictive accuracy. This work highlights substantial improvements in predictive accuracy and computational efficiency, advancing epidemiological forecasting and supporting real-time public health decision-making and AI-driven disease surveillance systems.

Task-oriented dialogue (TOD) system is designed to accomplish user-defined tasks through dialogues. The TOD system has progressed towards end-to-end modeling by leveraging pre-trained large language models. Fine-tuning the pre-trained language models using only supervised learning leads to the exposure bias and token loss problem and it deviates the models from completing the user’s task. To address these issues, we propose a TOD system that leverages a unified pre-trained language model, GPT-2, as a base model. It is optimized using supervised learning and offline reinforcement learning (RL). The issues in the TOD system are mitigated using a non-differentiable reward function. The reward is calculated using the weighted sum of the success rate and BLEU evaluation metrics. The success rate and BLEU metrics in reward calculation guide the language model for user task completion while ensuring a coherent and fluent response. Our model is acquired by fine-tuning a pre-trained model on the dialogue-session level which comprises user utterance, belief state, system act, and system response. Experimental results on MultiWOZ2.1 demonstrate that our model increases the inform rate by 1.60% and the success rate by 3.17% compared to the baseline.

Hierarchical Text Classification (HTC) aims to cate-gorize text data based on a structured label hierarchy, generating predicted labels that form a local hierarchical structure. Previous approaches have employed various methods to integrate text and label semantics, but they often overlooked the importance of local hierarchical context. By considering the local hierarchy, which encapsulates relationships between labels within the context of individual samples, we can enhance the association between text and its related labels. To address this, we propose a Margin Separation Loss (MSL), which explicitly models text-label semantic associations in a local hierarchy-aware manner. We obtain positive labels for each sample using the local hierarchy and employ the global hierarchy to identify corresponding negative labels. Positive and negative pairs are created by pairing the text sample with its positive and negative labels. MSL enforces a margin between positive and negative pairs at each hierarchical level, which ensures that similarity within positive pairs is maximized while similarity within negative pairs is minimized in the embedding space, thereby aligning text representations with their related labels. Building upon this, we introduce the Hierarchical Text-Label Association$(\mathbf{HTLA}^{\mathrm{n}})$model, utilizing BERT for text encoding and a customized Graphormer to encode label hierarchy and fusion of text-label embeddings to generate composite representations. Experimental results on benchmark datasets and comparison with existing baselines demonstrate the effectiveness of HTLAnfor HTC.

Accurate forecasting of water quality is vital for safeguarding public health, aquatic ecosystems and ensuring economic stability. This study introduces a novel hybrid approach using two-stage decomposition combined with Bidirectional Long Short-Term Memory and Attention mechanism (BiLSTM-Attention) for short-term forecasting of Dissolved Oxygen (DO). First, Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) decomposes the original data into a set of Intrinsic Mode Functions (IMFs) and residual component. Variational Mode Decomposition(VMD) further decomposes the IMF with the highest frequency into a set of modes. Then, BiLSTM-Attention is employed to generate the final forecast. The existing studies predominantly use MSE or MAE as the loss function for water quality forecasting. We design a customized dynamically weighted Huber loss function (DWHL) to optimize the training of our proposed forecasting model. The model is tested on data from ten locations along the river Ganga, consid-ering the forecasting horizons of one, two, and three hours, and compare it with ten models, both with and without decomposition. The results of the Diebold Mariano test suggests that there is a statistically significant difference in forecast accuracy of the proposed and compared models. The outcomes demonstrate that the proposed CEEMDAN-VMD-DWHL-BiLSTM-Attention approach effectively identifies complex and non-linear patterns in the data, resulting in significantly improved forecasts.

Intelligent data placement in data center networks for big data applications is a constructive approach to reduce the data movement delay, which can further improve the user service latency. Traditionally, data placement methods have relied on the study of the prior distribution of data requests discovered via network trace data analysis. Data placement in the network is an online decision-making problem. Despite the available machine learning architectures, how to deal with dynamic network conditions and user requests that vary over time is still an unanswered question. Toward this goal, this research formulates a Markov Decision Process (MDP) model based on deep reinforcement learning (RL) that takes into account the delay associated with data movement and service latency. The proposed method presents a novel intelligent data placement approach that leverages two state-of-the-art RL techniques; Deep Q-Network (DQN) and Soft Actor-Critic (SAC). Additionally, a prioritized butter is incorporated to replay important transitions more frequently and therefore, learn more efficiently. Extensive experiments and analyses conducted on Microsoft Research Cambridge (MSR) traces show the effectiveness and superior performance of the proposed approach for minimizing overall user service latency. The results highlight that our model outperforms the existing techniques in terms of the overall reward, achieving optimal data placement in real-world scenarios.

Effective text and label representations are crucial for accurate predictions in Hierarchical Text Classification (HTC). However, existing methods face challenges in capturing relevant label representations due to the discrete nature of labels and the complexities of their hierarchical relationships. To address these challenges, we propose Label Hierarchy Alignment (LHA), which ensures that learned label representations conform to the hierarchical label space. In this paper, we present two LHA approaches. The first is adversarial label alignment, which utilizes adversarial learning to enforce a prior distribution on label embeddings. This guides the model to adjust label embeddings in accordance with the hierarchical structure. The second approach is contrastive label alignment, which uses contrastive learning to identify semantic similarities and differences between label nodes across the hierarchy. This enriches the label embeddings with semantic depth, thereby enhancing their alignment with the hierarchical structure. By explicitly addressing various hierarchical aspects, both approaches generate refined label representations that align more effectively with the label hierarchy. We implemented LHA on the current state-of-the-art model and evaluated its performance on two benchmark datasets. The experiment results show that performing LHA improves HTC performance.

The COVID-19 pandemic has resulted in widespread illness and death around the world. Massive and complex Electronic health records (EHRs) hold important information to help COVID-19 patients get an accurate diagnosis and prognosis. In this research, we suggested a novel ensemble of decision trees method for analyzing and classifying COVID-19 EHR data. The proposed approach uses misclassified samples at each iteration to improve accuracy. Pneumonia, hypertension, diabetes, male gender, and higher age groups were the most important features in classifying patients into “COVID-negative” “COVID-positive and survived” and “COVID-positive and died” To evaluate the efficacy of our suggested ensemble model, we contrasted it with well-established models. Compared to the baseline models, the experimental findings demonstrated that the ensemble model greatly outperformed in terms of accuracy, precision, recall, and F1 score. This research shows the significance of EHRs in COVID-19 patient classification and suggests an innovative method to improve classification accuracy. The most prominent traits in high-risk patients can serve as early warning indicators, allowing for quick intervention and better outcomes overall. The suggested ensemble model has practical applications in the early detection and efficient patient treatment of COVID-19.

This paper summarizes the Crowdsensing-based Road Damage Detection Challenge (CRDDC), a Big Data Cup organized as a part of the IEEE International Conference on Big Data’2022. The Big Data Cup challenges involve a released dataset and a well-defined problem with clear evaluation metrics. The challenges run on a data competition platform that maintains a real-time online evaluation system for the participants. In the presented case, the data constitute 47,420 road images collected from India, Japan, the Czech Republic, Norway, the United States, and China to propose methods for automatically detecting road damages in these countries. More than 70 teams from 19 countries registered for this competition. The submitted solutions were evaluated using five leaderboards based on performance for unseen test images from the aforementioned six countries. This paper encapsulates the top 11 solutions proposed by these teams. The best-performing model utilizes ensemble learning based on YOLO and Faster-RCNN series models to yield an F1 score of 76% for test data combined from all 6 countries. The paper concludes with a comparison of current and past challenges and provides direction for the future.

Federated Learning (FL) is a distributed training framework where a model is collaboratively trained over a set of clients without communicating their private data to the central server. However, each client shares the parameters of its local model. The first challenge faced by the FL is high communication cost due to the size of Deep Neural Network (DNN) models. Pruning is an efficient technique to reduce the number of parameters in DNN models, in which insignificant neurons are removed from the model. This paper introduces a federated pruning method based on Neuron Importance Scope Propagation (NISP) algorithm. The importance scores of output layer neurons are back-propagated layer-wise to every neuron in the network. The central server iteratively broadcasts the sparsified weights to all selected clients. Then, each participating client intermittently downloads the mask vector and reconstructs the weights in their original form. The locally updated model is pruned using the mask vector and shared with the server. After receiving model updates from each participating client, the server reconstructs and aggregates the weights. Experiments on MNIST and CIFAR10 datasets demonstrate that the proposed approach achieves accuracy close to Federated Averaging (FedAvg) algorithm with less communication cost.

This paper summarizes the Global Road Damage Detection Challenge (GRDDC), a Big Data Cup organized as a part of the IEEE International Conference on Big Data’2020. The Big Data Cup challenges involve a released dataset and a well-defined problem with clear evaluation metrics. The challenges run on a data competition platform that maintains a leaderboard for the participants. In the presented case, the data constitute 26336 road images collected from India, Japan, and the Czech Republic to propose methods for automatically detecting road damages in these countries. In total, 121 teams from several countries registered for this competition. The submitted solutions were evaluated using two datasets test1 and test2, comprising 2,631 and 2,664 images. This paper encapsulates the top 12 solutions proposed by these teams. The best performing model utilizes YOLO-based ensemble learning to yield an F1 score of 0.67 on test1 and 0.66 on test2. The paper concludes with a review of the facets that worked well for the presented challenge and those that could be improved in future challenges.

In February 2020, COVID-19 was declared an infodemic by the experts in the situational report published by WHO. A study by the Reuters Institute for the Study of Journalism found that 40% believed that social media is the primary source of misinformation. Some research works have focused on analyzing the misinformed Twitter data using bot detection and fake-URL identification tools. However, these methods require annotated data in advance and high capacity hardware resources to deal with vast amounts of data. To overcome this challenge, we propose an unsupervised framework to detect misinformed content and the users who may be the sources of or susceptible to spreading misinformed content. This research work leverages the popularity of the tweets in form of re-posts to identify the influential content on the social media. A long w ith textual information, the credibility of the original posters of the popular tweets is assessed by using four primary credible sources i.e., News, Medical Organization, Medical Influencers, a nd Verified users. These categories are dynamically updated by using the semantic similar textual information from the user-descriptions of these user categories. The framework has been tested on two public annotated datasets from FakeNewsNet and has been compared with the baseline classification, clustering, and hybrid models for misinformation detection. Due to the unavailability of the annotated Twitter data, the resulting suspected misinformed users were searched manually on Twitter, where approximately 50% of identified users were a lready suspended. The content of misinformed clusters was validated by performing an emotional analysis of the tweets. This showed that the suspected misinformed clusters contained a high affinity towards anger and fear-inducing tweets. Furthermore, we analyzed the temporal change in geo-graphical tweet densities for all the user-categories and the popular user-interactions in the whole data.

Frequent itemset mining is used to extract interesting associations and correlations between the itemsets present in transactional datasets. The frequently appearing patterns are used for various business decision making policies, for instance to increase co-purchase of products, price optimization, cross promotion etc. However, there are some sensitive patterns present in datasets that can reveal individual or organisation's specific confidential information that they would not prefer to be known since it can cause them huge social and monetary loss. Privacy Preserving Data Mining (PPDM) approaches are used to hide these sensitive patterns with maintaining the utility of the data. Heuristics-based PPDM approaches are widely adopted sensitive pattern hiding approaches due to their simplicity and lesser computational time as compared to the border-based and exact approaches. However, these approaches causes high side effects concerning the quality of datasets. In this paper, two heuristics-based algorithms, Removal of Closed Sensitive Itemsets with Maximum Support (MaxRCSI) and Removal of Closed Sensitive Itemsets with Minimum Support (MinRCSI), are proposed. In these algorithms, data sanitization is performed over closed sensitive itemsets to improve the utility of sanitized data. The proposed algorithms are parallelized on Spark parallel computing framework to deal with the massive amount of data i.e. big data. Experiments performed on real and synthetic datasets show that the proposed algorithms preserve the privacy of datasets with substantially better utility as compared to the traditional algorithms with less execution time.