ABSTRACT Forecasting cryptocurrency prices remains challenging due to extreme volatility, regime‐dependent dynamics, and unstable cross‐asset correlations. Statistical methods such as ARIMA and GARCH assume stationarity and linear dependence structures, making them inadequate for capturing non‐linear temporal patterns in high volatile cryptocurrency data. Conventional machine learning methods often require hand‐crafted features and fail to capture the sequential temporal dependencies inherent in price series. Recurrent deep learning approaches such as recurrent neural networks (RNNs) and LSTMs address the issues but suffer from limited parallelization, vanishing gradients, and difficulty in learning multi‐scale temporal patterns. The advances in Transformer have demonstrated strong capability in capturing long‐range temporal dependencies through self‐attention while enabling parallel computation. However, canonical Transformer still struggle with noisy, volatile financial time series due to computational complexity and sensitivity to irrelevant temporal patterns. These limitations motivate the exploration of Transformer‐based architectures. Our study makes three key contributions. First, we propose MSPCIFormer, a novel Transformer‐based architecture that integrates multi‐scale patching with channel‐independent (CI) modelling to capture heterogeneous temporal dynamics while mitigating noise from time‐varying inter‐asset correlations. Second, we conduct comprehensive experiments comparing MSPCIFormer with state‐of‐the‐art Transformer models and strong time‐series forecasting baselines, evaluated using both statistical and economic metrics across multiple forecasting horizons. Third, we establish a unified evaluation framework incorporating both Hold‐out and Walk‐forward evaluation framework with economical metrics for regime‐robustness testing. Empirical results demonstrate that MSPCIFormer achieves the best or tied‐best predictive accuracy among all Transformer‐based baselines across three cryptocurrency assets, while maintaining competitive and stable performance across diverse market regimes.

Association rule mining techniques can generate a large volume of sequential data when implemented on transactional databases. Extracting insights from a large set of association rules has been found to be a challenging process. When examining a ruleset, the fundamental question is how to summarise and represent meaningful mined knowledge efficiently. Many algorithms and strategies have been developed to address issue of knowledge extraction; however, the effectiveness of this process can be limited by the data structures. A better data structure can sufficiently affect the speed of the knowledge extraction process. This paper proposes a novel data structure, called the Trie of rules, for storing a ruleset that is generated by association rule mining. The resulting data structure is a prefix-tree graph structure made of pre-mined rules . This graph stores the rules as paths within the prefix-tree in a way that similar rules overlay each other. Each node in the tree represents a rule where a consequent is this node, and an antecedent is a path from this node to the root of the tree. The evaluation showed that the proposed representation technique shows significant value. It compresses a ruleset with no data loss and benefits in terms of time for basic operations such as searching for a specific rule, which is the base for many knowledge discovery methods. Moreover, our method demonstrated a significant improvement in graph traversal time compared to traditional data structures.

Cryptocurrencies have won a lot of attention as an investment tool in recent years. Specific research has been done on cryptocurrencies’ price prediction while the prices surge up. Classic models and recurrent neural networks are applied for the time series forecast. However, there remains limited research on how the Transformer works on forecasting cryptocurrencies price data. This paper investigated the forecasting capability of the Transformer model on Bitcoin (BTC) price data and Ethereum (ETH) price data which are time series with high fluctuation. Long short term memory model (LSTM) is employed for performance comparison. The result shows that LSTM performs better than Transformer both on BTC and ETH price prediction. Furthermore, in this paper, we also investigated if sentiment analysis can help improve the model’s performance in forecasting future prices. Twitter data and Valence Aware Dictionary and sEntiment Reasoner (VADER) is used for getting sentiment scores. The result shows that the sentiment analysis improves the Transformer model’s performance on BTC price but not ETH price. For the LSTM model, the sentiment analysis does not help with prediction results. Finally, this paper also shows that transfer learning can help on improving the Transformer’s prediction ability on ETH price data.

Due to the COVID19 pandemic, more higher-level education programmes have moved to online channels, raising issues in monitoring students’ learning progress. Thanks to advances in online learning systems, however, student data can be automatically collected and used for the investigation and prediction of the students’ learning performance. In this article, we present a novel approach to analyse students’ learning behaviour, as well as the relationship between these behaviours and learning assessment results, in the context of programming education. A bespoke method has been built based on a combination of Random Matrix Theory, a Community Detection algorithm and statistical hypothesis tests. The datasets contain fine-grained information about students’ learning behaviours in two programming courses over two academic years with about 400 first-year students in a Medium-sized Metropolitan University in Dublin. The proposed method is a noval approach to data preprocessing which can improve the analysis and prediction based on learning behavioural datasets. The proposed approach deals with the issues of noise and trend effect in the data and has shown its success in detecting groups of students who have similar learning behaviours and outcomes. The higher performing groups have been found to be more active in practical-related activities throughout the course. Conversely, we found that the lower performing groups engage more with lecture notes instead of doing programming tasks. The learning behaviours data can also be used to predict students’ outcomes (i.e. Pass or Fail the terminal exams) at the early stages of the study, using popular machine learning classification techniques.

Sentiment Analysis techniques enable the automatic extraction of sentiment in social media data, including popular platforms as Twitter. For retailers and marketing analysts, such methods can support the understanding of customers' attitudes towards brands, especially to handle crises that cause behavioural changes in customers, including the COVID-19 pandemic. However, with the increasing adoption of black-box machine learning-based techniques, transparency becomes a need for those stakeholders to understand why a given sentiment is predicted, which is rarely explored for retailers facing social media crises. This study develops an Explainable Sentiment Analysis (XSA) application for Twitter data, and proposes research propositions focused on evaluating such application in a hypothetical crisis management scenario. Particularly, we evaluate, through discussions and a simulated user experiment, the XSA support for understanding customer's needs, as well as if marketing analysts would trust such an application for their decision-making processes. Results illustrate the XSA application can be effective in providing the most important words addressing customers sentiment out of individual tweets, as well as the potential to foster analysts' confidence in such support.

Received total wideband power (RTWP) data is a measurement of the wanted and unwanted power levels received by a 3G radio base station (RBS) and is a concise indicator of uplink network performance. Using a statistical physics approach, we aim to detect periods of unusual activity between cells by assessing a sample of RTWP measurement data from a live network. Using wavelet correlation and cross-correlation techniques we analyse multivariate non-stationary time series for statistical relationships at different time scales. We analyse the seasonal component of the dataset as well as examining the autocorrelation and partial autocorrelation methods. We then explore the Hurst exponent of the dataset and inspect the intraday correlations for patterns of events. Next, we examine the eigenvalue spectrum using different sized sliding windows. Finally, we compare approaches for assessing multiscale relationships among several variables using the wavelet multiple correlation and wavelet zero-lag cross-correlation on non-stationary RTWP time series data.

Predicting currency prices remains a difficult endeavour. Investors are continually seeking new ways to extract\nmeaningful information about the future direction of price changes. Recently, cryptocurrencies have attracted\nhuge attention due to their unique way of transferring value as well as its value as a hedge. A method proposed\nin this project involves using data mining techniques: mining text documents such as news articles and tweets\ntry to infer the relationship between information contained in such items and cryptocurrency price direction.\nThe Long Short-Term Memory Recurrent Neural Network (LSTM RNN) assists in creating a hybrid model\nwhich comprises of sentiment analysis techniques, as well as a predictive machine learning model. The success\nof the model was evaluated within the context of predicting the direction of Bitcoin price changes. Findings\nreported here reveal that our system yields more accurate and real-time predictions of Bitcoin price fluctuations\nwhen compared to other existing models in the market.

Features or functionalities provided by cloud-based applications are accessed by users through various interfaces \nsuch as web browser, mobile app, and command line interface. Yet for monitoring cloud-based applications, \nsoftware developers and researchers have focused on web browsers. Software updates are provided \nfor such applications based on the data acquired from the cloud monitoring components but usage data of the \ncloud application features are difficult to extract in a cloud environment as the usage data is spread across the \ninterfaces on the front-end and the back-end. In this paper, we focus on the usage of the cloud application \nfeatures from the user perspective and how to extract these data in a cloud environment. We define six criteria \nfor the user-level usage data, analyse the existing usage data extraction techniques and propose a usage data \nextraction framework adhering to the defined criteria.

The extremely large number of devices available to the modern day user, with the increase in device intercommunication, is fuelling the latest Internet of Things (IoT) development. IoT needs to enable exchange of various types of data, from sensor data to multimedia, between numerous diverse devices differing in power, connectivity, mobility and energy, while also maintaining high levels of Quality of Service (QoS). This paper performs statistical analysis of the innovative NETworking Scheme for sMart IoT gatewayS (NETSMITS) in terms of several QoS network-related metrics with most significant impact on devices' performance. NETSMITS introduces an innovative algorithm which uses QoS and service relevance metrics in order to efficiently cluster intercommunicating IoT objects. Statistical analysis is performed on the QoS data collected in a highly relevant multi-device scenario in order to understand NETSMITS' behaviour. Interesting results were obtained, describing the relationship between the QoS metrics and different types of IoT devices.

One of the main challenges in data analytics is that discovering structures and patterns in complex datasets is a computer-intensive task. Recent advances in high-performance computing provide part of the solution. Multicore systems are now more affordable and more accessible. In this paper, we investigate how this can be used to develop more advanced methods for data analytics. We focus on two specific areas: model-driven analysis and data mining using optimisation techniques.