Search

The main problem in building intelligent systems is the lack of data for machine learning, which is especially important for sign language recognition for the deaf and hard of hearing. One of the ways to increase the amount of data for training is synthesis. Unlike speech synthesis, it is impossible to create a sequence of gestures in Vietnamese and some other languages that exactly repeat the text. This is due to the significant limitations of the gesture dictionary and the different word order in sentences. The aim of the work is to enrich the educational corpus of video data for use in creating recognition systems for the Vietnamese Sign Language (ViSL). Since it is impossible to translate the words of the source text into gestures one to one, the problem of translating from a regular language into a sign language arises. The paper proposes to use a two-phase process for this. The first phase involves pre-processing the text with standardization of the text format, segmentation of words and sentences, and then encoding the words using the sign language dictionary. At this stage, it should be noted that there is no need to remove punctuation marks and stop words, since they are related to the accuracy of the N-gram model. Next, instead of using syntactic analysis, a statistical method for forming a sequence of gestures is used, and the Markov model on the transition graph between words is taken as a basis in which the probability of the next word depends only on the two previous words. Transition probabilities are calculated on the existing marked corpus of the ViSL. The Breadth-first Search method is used to compile a list of all sentences generated based on a given grammatical rule and a matrix of semantic interactions between words. The inverse of the logarithm of the product of the probabilities of co-occurrence of consecutive 3-word phrases in a sentence is used to estimate the frequency of occurrence of that sentence in a given data set. Based on the ViSL data of 3,234 words, we calculated probability matrices representing the relationships between words based on Vietnamese natural language data with 50 million sentences collected from Vietnamese newspapers and magazines. For different grammar rules, we compare the number of generated sentences and evaluate the accuracy of the 50 most frequent sentences. The average accuracy is 88 %. The accuracy of the generated sentences is estimated by manual statistical methods. The number of generated sentences depends on the number of word parts that are labeled according to the grammar rules. The semantic accuracy of the generated sentences will be very high if the search words are labeled with the correct part-of-speech tagging. Compared with machine learning methods, our proposed method gives very good results for languages without inflections and word order that follow certain rules, such as Vietnamese, and does not require large computational resources. The disadvantage of this method is that its accuracy largely depends on the type of word, sentence, and word segmentation. The relationship of words depends on the observed dataset. Future research direction is to generate paragraphs in sign language. The obtained data can be used in machine learning models for sign language processing tasks.

This paper presents and discusses a framework that leverages large-scale language models (LLMs) for data enrichment and continuous monitoring emphasizing its essential role in optimizing the performance of deployed models. It introduces a comprehensive large language model operations (LLMOps) methodology based on continuous monitoring and continuous improvement of the data, the primary determinant of the model, in order to optimize the prediction of a given phenomenon. To this end, first we examine the use of real-time web scraping using tools such as Kafka and Spark Streaming for data acquisition and processing. In addition, we explore the integration of LLMOps for complete lifecycle management of machine learning models. Focusing on continuous monitoring and improvement, we highlight the importance of this approach for ensuring optimal performance of deployed models based on data and machine learning (ML) model monitoring. We also illustrate this methodology through a case study based on real data from several real estate listing sites, demonstrating how MLflow can be integrated into an LLMOps pipeline to guarantee complete development traceability, proactive detection of performance degradations and effective model lifecycle management. [ABSTRACT FROM AUTHOR]

Emergency services rely heavily on Twitter for early detection of crisis tasks to enhance crisis management systems. However, employing state-of-the-art models often face data sparsity as well as their inadequacy to handle long-range dependencies between tweet tokens. Additionally, the authorities need to gain confidence in the model's prediction so that the detected task information can be better believed and prioritized. In this study, we present a generalized framework named explainable attentive model for crisis task identification (ExACT) to handle the above mentioned challenges, while identifying crisis task relevant tweets as well as provide the model explainability by utilizing a very small corpus of tweets. The novelty of ExACT is two-fold: (1) Data enrichment has been introduced by nondynamic contextual attributes derived from tweets to overcome the sparsity and improve data quality. (2) Feature enrichment has been incorporated using hierarchical attention at both local and global levels using residual self-attention and correlation attention to capture long-range dependencies. Additionally, LIME based explainability approach added to understand the task important tokens. Experiments reveal that ExACT has a competitive performance improvement over various state-of-the-art models in terms of F 1 -score ( 20 % and 14 % respectively) and accuracy ( 14 % and 16 % , respectively) across two different crisis tasks infrastructure damage and support signal identification. Consistent performance improvement for two different tasks considered from publicly available crisis event datasets depicts the model's generalizability. While, LIME supported explainable mechanism in ExACT can identify the important keywords but does not guarantee a high score in terms of plausibility and faithfulness metrics. [ABSTRACT FROM AUTHOR]

Rare events are occurrences that take place with a significantly lower frequency than more common, regular events. These events can be categorized into distinct categories, from frequently rare to extremely rare, based on factors like the distribution of data and significant differences in rarity levels. In manufacturing domains, predicting such events is particularly important, as they lead to unplanned downtime, a shortening of equipment lifespans, and high energy consumption. Usually, the rarity of events is inversely correlated with the maturity of a manufacturing industry. Typically, the rarity of events affects the multivariate data generated within a manufacturing process to be highly imbalanced, which leads to bias in predictive models. This paper evaluates the role of data enrichment techniques combined with supervised machine learning techniques for rare event detection and prediction. We use time series data augmentation and sampling to address the data scarcity, maintaining its patterns, and imputation techniques to handle null values. Evaluating 15 learning models, we find that data enrichment improves the F1 measure by up to 48% in rare event detection and prediction. Our empirical and ablation experiments provide novel insights, and we also investigate model interpretability. [ABSTRACT FROM AUTHOR]

In this study, two innovative techniques were introduced, including data enrichment and optimization, with the aim of significantly improving the accuracy of air over-pressure (AOP) prediction models in mine blasting. Firstly, the Extra Trees algorithm was applied to enrich the collected dataset with the goal of enhancing the understanding of the predictive models for AOP prediction. Then, a neural network model with two hidden layers (ANN) was designed to predict AOP using both the original and enriched datasets. Secondly, to further enhance the accuracy of the ANN model, a novel optimization algorithm based on a random walk strategy and the grey wolf optimization algorithm (RWGWO) was employed to optimize the weights of the ANN model. This optimized model, referred to as the RWGWO–ANN model, was developed and evaluated for predicting AOP using both the original and enriched datasets. To comprehensively assess the impact of data enrichment and the proposed RWGWO-ANN model, three other optimization algorithms—particle swarm optimization (PSO), fruit-fly optimization algorithm (FOA), and single-based genetic algorithm (SGA)—were also applied to optimize the ANN model for AOP prediction. These models were named PSO–ANN, FOA–ANN, and SGA–ANN, respectively. The tenfold cross-validation procedure was applied and repeated three times to ensure the objectivity and consistency of the models. Additionally, conventional ANN and the United States Bureau of Mines empirical model were developed for comparison, serving similar purposes to evaluate the efficiency of the optimization algorithms employed in this study. To demonstrate the advantages of the proposed method and models, a dataset comprising 312 blasting events and six input parameters at the Coc Sau open-pit coal mine in Vietnam was gathered and analyzed. These parameters included burden, spacing, rock hardness, powder factor, monitoring distance, and maximum explosive charge per delay. An additional input variable—Extra Trees—was introduced, making the total number of input variables seven in the enriched dataset. The proposed hybrid model, along with others, was developed based on both the original and enriched datasets. The results revealed that the Extra Trees algorithm is robust and effectively enriches the raw dataset, enhancing the understanding of predictive models and providing improved accuracy. Sensitivity analysis results also highlighted the robust contribution of the Extra Trees variable in the enriched dataset. Compared to the original dataset, the performance of AOP predictive models was improved by 7–24% using the enriched dataset enriched by the Extra Trees algorithm. Furthermore, the findings indicated that the RWGWO–ANN model exhibited the highest accuracy in predicting AOP in this study, achieving an accuracy of 96.2%. This marked a 16–20% improvement over the accuracy of the conventional ANN model. [ABSTRACT FROM AUTHOR]

The vast amount of information stemming from the deployment of the Internet of Things and open data portals is poised to provide significant benefits for both the private and public sectors, such as the development of value-added services or an increase in the efficiency of public services. This is further enhanced due to the potential of semantic information models such as NGSI-LD, which enable the enrichment and linkage of semantic data, strengthened by the contextual information present by definition. In this scenario, advanced data processing techniques need to be defined and developed for the processing of harmonised datasets and data streams. Our work is based on a structured approach that leverages the principles of linked-data modelling and semantics, as well as a data enrichment toolchain framework developed around NGSI-LD. Within this framework, we reveal the potential for enrichment and linkage techniques to reshape how data are exploited in smart cities, with a particular focus on citizen-centred initiatives. Moreover, we showcase the effectiveness of these data processing techniques through specific examples of entity transformations. The findings, which focus on improving data comprehension and bolstering smart city advancements, set the stage for the future exploration and refinement of the symbiosis between semantic data and smart city ecosystems. [ABSTRACT FROM AUTHOR]

In the big data context, it is very frequent to manage the analysis of missing values. This is especially relevant in the field of statistical analysis, where this represents a thorny issue. This study proposes a strategy for data enrichment in presence of sparse matrices. The research objective consists in the evaluation of a possible distinction of behaviour among observations in sparse matrices with missing data. After selecting among the multiple imputation methods, an innovative technique will be presented to impute missing observations as a negative position or a neutral opinion. This method has been applied to a dataset measuring the interaction between users and social network pages for some Italian newspapers. [ABSTRACT FROM AUTHOR]

Making decisions about which objects to keep or omit is challenging in map design. This process, called selection, constitutes the first operation in cartographic generalization. In this research, a method of automatic road selection for creating small-scale maps using machine learning and data enrichment is proposed. First, the problem of contextual information scarcity concerning roads in the source database is addressed. Additional information concerning the relations between roads and other objects was added (such as centrality and proximity measures). Second, machine learning is used to design automatic selection models based on enriched information. Third, three different road selection approaches are implemented. The baseline approach is following the official map design guidelines. The second approach is based on machine learning using the enriched road database. The third approach is based on an existing structural model. The results of all approaches are compared to existing atlas maps designed by experienced cartographers. The results of the Machine Learning Approaches were most similar to the atlas maps (between 81% and 90% accuracy). The least efficient approaches were the Structural Approach with 32% and the Guidelines Approach with 44% accuracy. We conclude that enriching road data with new contextual information concerning roads and using machine learning is beneficial as the achieved results outperform both Guidelines and Structural Approaches. [ABSTRACT FROM AUTHOR]

Ricgraph, also known as Research in context graph, enables the exploration of researchers, teams, their results, collaborations, skills, projects, and the relations between these items.Ricgraph can store many types of items into a single graph. These items can be obtained from various systems and from multiple organizations. Ricgraph facilitates reasoning about these items because it infers new relations between items, relations that are not present in any of the separate source systems. Ricgraph is flexible and extensible, and can be adapted to new application areas.In this article, we illustrate how Ricgraph works by applying it to the application area research information.

Energy simulation models are crucial to estimate the energy demand of buildings, especially for prospective planning on a district or city scale. As required input data is not available in many cases, an automated model generation workflow is needed. Existing workflows have several disadvantages, including: (i) dependence on large input datasets of existing buildings; (ii) no 3D representation to support the planning process; (iii) they are proprietary solutions. The pipeline ‘SHP2SIM’ is an open-source python pipeline enabling enrichment and generation of building energy simulation models based on little input data for district and urban scale. The pipeline is tested by simulating the heat load for a district with 27 buildings and validated for one building: R squared is 0.9825, CV(RMSE) is 22.10%, and NMBE is 4.06% on a monthly basis. To enable reproducibility and encourage open science, input data, output models, and the pipeline are openly available (https://github.com/tug-cps/shp2sim).

Rare events are occurrences that take place with a significantly lower frequency than more common, regular events. These events can be categorized into distinct categories, from frequently rare to extremely rare, based on factors like the distribution of data and significant differences in rarity levels. In manufacturing domains, predicting such events is particularly important, as they lead to unplanned downtime, a shortening of equipment lifespans, and high energy consumption. Usually, the rarity of events is inversely correlated with the maturity of a manufacturing industry. Typically, the rarity of events affects the multivariate data generated within a manufacturing process to be highly imbalanced, which leads to bias in predictive models. This paper evaluates the role of data enrichment techniques combined with supervised machine learning techniques for rare event detection and prediction. We use time series data augmentation and sampling to address the data scarcity, maintaining its patterns, and imputation techniques to handle null values. Evaluating 15 learning models, we find that data enrichment improves the F1 measure by up to 48% in rare event detection and prediction. Our empirical and ablation experiments provide novel insights, and we also investigate model interpretability.

Energy simulation models are crucial to estimate the energy demand of buildings, especially for prospective planning on a district or city scale. As required input data is not available in many cases, an automated model generation workflow is needed. Existing workflows have several disadvantages, including: (i) dependence on large input datasets of existing buildings; (ii) no 3D representation to support the planning process; (iii) they are proprietary solutions. The pipeline 'SHP2SIM' is an open-source python pipeline enabling enrichment and generation of building energy simulation models based on little input data for district and urban scale. The pipeline is tested by simulating the heat load for a district with 27 buildings and validated for one building: R squared is 0.9825, CV(RMSE) is 22.10%, and NMBE is 4.06% on a monthly basis. To enable reproducibility and encourage open science, input data, output models, and the pipeline are openly available (). [ABSTRACT FROM AUTHOR]

This study examines the leverage of big data to enrich a toponymy repository with the objective to facilitate the spatial–historical investigation of toponymy in Bali. The strong Hinduism cultural background becomes the sacred values that generally affect the toponymic formation. The dataset constitutes historic maps from Dutch Royal Tropical Institute digital library and online databases from several sources such as GeoNames, OpenStreetMap, and Indonesian Geospatial Information Agency. The mixed method leverages the data enrichment technique with descriptive and historical analysis to strengthen the discourse of toponymy formation. The geographical variables to enrich the toponyms data are including elevation, geomorphology, land cover, and rivers, prepared in shapefiles. Some technical procedures are conducted to create a geodatabase such as georeferentiation, digitisation, data manipulation and cleaning, and overlay analysis resulting in 16,923 clean toponyms data entries. The overlay analysis indicates different trends in each class of geographic variables. The functional-semantic and lexical–morphological basis of name-giving defines name formation in correlation with the name models that examine the linguistic elements. Some main groups of place names and their etymologies are identified such as Banjar, Gunung, Tukad, Padang, Bukit, and Gili are associated with geomorphological factors such as volcanic formations, freshwaters, and marine. Further explorations are suggested to dig deeper into the discourse of toponymy in Bali as a complex system involves the typical landscapes, cultural as well as political and social characteristics of the region. [ABSTRACT FROM AUTHOR]

The issue of blast-induced ground vibration poses a significant environmental challenge in open-pit mines, necessitating precise prediction and control measures. While artificial intelligence and machine learning models hold promise in addressing this concern, their accuracy remains a notable issue due to constrained input variables, dataset size, and potential environmental impact. To mitigate these challenges, data enrichment emerges as a potential solution to enhance the efficacy of machine learning models, not only in blast-induced ground vibration prediction but also across various domains within the mining industry. This study explores the viability of utilizing machine learning for data enrichment, with the objective of generating an augmented dataset that offers enhanced insights based on existing data points for the prediction of blast-induced ground vibration. Leveraging the support vector machine (SVM), we uncover intrinsic relationships among input variables and subsequently integrate them as supplementary inputs. The enriched dataset is then harnessed to construct multiple machine learning models, including k-nearest neighbors (KNN), classification and regression trees (CART), and random forest (RF), all designed to predict blast-induced ground vibration. Comparative analysis between the enriched models and their original counterparts, established on the initial dataset, provides a foundation for extracting insights into optimizing the performance of machine learning models not only in the context of predicting blast-induced ground vibration but also in addressing broader challenges within the mining industry [ABSTRACT FROM AUTHOR]

In recent years, the Building Information Modeling (BIM) has been the subject of extensive scientific literature. However, its application in historical construction still requires theoretical research and experimentation. The primary drawback is linked to the fact that historical heritage does not align with the standardization principles inherent in BIM modeling, which is predominantly tailored for new construction endeavors. The application of BIM modeling to the Oratory of S. Giovanni Battista in Bussana Vecchia, Sanremo (Imperia) aims to conduct a critical analysis of the methodology by creating an intelligent parametric 3D model capable of containing all the information necessary for an executive restoration project of the object. [ABSTRACT FROM AUTHOR]

Proliferation of data sources associated to Internet of Things (IoT) deployment as well as those bound to Open Data Portals (e.g. European Data Portal, Municipalities Open Data Portals, etc.) and Social Media platforms is creating an abundance of information that is called to bring benefits for both the private and public sectors, through the development of added-value services, increasing administrations’ transparency and availability or fostering efficiency of public services. However, pieces of information without a context are significantly less valuable. Raw data lacks semantics and it is highly heterogeneous from one data-source to another. This poses a challenge to make it useful. To turn all this data into valuable information it is necessary to enable its combination so that meaningful context can be created. Moreover, it is fundamental to define the mechanisms enabling the adoption and orchestration of advanced (typically AI-enabled) data processing techniques to be applied over the harmonized datasets and data-streams. This paper presents the Data Enrichment Toolchain (DET) that provides the necessary harmonization and enrichment to datasets and data-streams coming from heterogeneous sources. The value of the enriched data lies on the one hand in the transfer of the data into a semantically grounded knowledge graph and, on the other hand, in the creation of new data through linking, aggregating and reasoning on the data. In both cases, the benefit of employing linked-data modelling and semantics comes from the extension of the metadata that is associated to every piece of information. Furthermore, the experimental evaluation of the DET implementation that we have carried out is also presented in the paper.

The shift towards renewable energy integration into smart grids has led to complex management processes, which require finer-grained energy and heat generation/ demand forecasting while considering data from monitoring devices and the integration of smaller multi-energy sub-systems at the community, district, or buildings level. However, energy prediction is challenging due to the high variability in the electrical and thermal energy demands of building occupants, the heterogenous characteristics of the energy assets or buildings in a district, and the length of the forecasting horizon. In this paper, we define a data-driven machine-learning pipeline to predict the electricity and thermal consumption of buildings and energy assets from a city district in 24 h intervals. Each pipeline's step is divided into sensors' data processing and model integration, data enrichment and features engineering, and multilayer perceptron model training. To address some of the drawbacks of using the multi-layer perceptron model, such as slow convergence rate and risk of overfitting, and to ensure a lower error in the energy prediction process features, an engineering technique was employed. We incorporated weather data features and interaction features derived from fusing the energy data with statistical models to capture the nonlinear patterns of the electrical and heat demands. The proposed approach was successfully validated in a real-world environment, a city district in Gent, Belgium. It featured good prediction results for electricity and heat production and consumption of various assets without considering the physical characteristics, making it viable and easily applicable in broader urban areas. The evaluation of energy prediction accuracy yielded good results, with a Mean Absolute Error (MAE) falling within the range of 0.003 to 3.27, and a Mean Absolute Scaled Error (MASE) ranging from 7 × 10−5 to 2.57 × 10−3. [ABSTRACT FROM AUTHOR]

Particulate matter (PM) in air has been proven to be hazardous to human health. Here we focused on analysis of PM data we obtained from the same campaign which was presented in our previous study. Multivariate linear and random forest models were used for the calibration and analysis. In our linear regression model the inputs were PM, temperature and humidity measured with low-cost sensors, and the target was the reference PM measurements obtained from SEPA in the same timeframe. [ABSTRACT FROM AUTHOR]

Multi-task learning in deep neural networks has become a topic of growing importance in many research fields, including drug discovery. However, applying multi-task learning poses new challenges in improving prediction performance. This study investigated the potential of training data enrichment to enhance multi-task model prediction quality in drug discovery. The study evaluated four scenarios with varying degrees of information capacity of the training data and applied two types of test data to evaluate prediction performance. We used three datasets: ViralChEMBL, which consisted of binary activities of compounds against viral species, was applied for the classification task; pQSAR(159) and pQSAR(4267), which consisted of bio-activities of compounds and assays from the research of the profile-QSAR method, were applied for regression tasks. We built multi-task models based on the feed-forward DNNs using the PyTorch framework. Our findings showed that training data enrichment could be an effective means of enhancing prediction performance in multi-task learning, but the degree of improvement depends on the quality of the training data. The more unique compounds and targets the training data included, the more new compound-target interactions are required for prediction improvement. Also, we found out that even using multi-task learning, one could not predict the interactions of compounds that are highly dissimilar from those used for model training. The study provides some recommendations for effectively employing multi-task learning in drug discovery to improve prediction accuracy and facilitate the discovery of novel drug candidates. [ABSTRACT FROM AUTHOR]

The vast amount of information stemming from the deployment of the Internet of Things and open data portals is poised to provide significant benefits for both the private and public sectors, such as the development of value-added services or an increase in the efficiency of public services. This is further enhanced due to the potential of semantic information models such as NGSI-LD, which enable the enrichment and linkage of semantic data, strengthened by the contextual information present by definition. In this scenario, advanced data processing techniques need to be defined and developed for the processing of harmonised datasets and data streams. Our work is based on a structured approach that leverages the principles of linked-data modelling and semantics, as well as a data enrichment toolchain framework developed around NGSI-LD. Within this framework, we reveal the potential for enrichment and linkage techniques to reshape how data are exploited in smart cities, with a particular focus on citizen-centred initiatives. Moreover, we showcase the effectiveness of these data processing techniques through specific examples of entity transformations. The findings, which focus on improving data comprehension and bolstering smart city advancements, set the stage for the future exploration and refinement of the symbiosis between semantic data and smart city ecosystems.

Wind power forecasting involves data preprocessing and modeling. In pursuit of better forecasting performance, most previous studies focused on creating various wind power forecasting models, but few studies have been published with an emphasis on new types of data preprocessing methods. Effective data preprocessing techniques and the fusion with the physical nature of the wind have been called upon as potential future research directions in recent reviews in this area. Data enrichment as a method of data preprocessing has been widely applied to forecasting problems in the consumer data universe but has not seen application in the wind power forecasting area. This study proposes data enrichment as a new addition to the existing library of data preprocessing methods to improve wind power forecasting performance. A methodological framework of data enrichment is developed with four executable steps: add error features of weather prediction sources, add features of weather prediction at neighboring nodes, add time series features of weather prediction sources, and add complementary weather prediction sources. The proposed data enrichment method takes full advantage of multiple commercially available weather prediction sources and the physical continuity nature of wind. It can cooperate with any existing forecasting models that have weather prediction data as inputs. The controlled experiments on three actual individual wind farms have verified the effectiveness of the proposed data enrichment method: The normalized root mean square error (NRMSE) of the day-ahead wind power forecast of XGBoost and LSTM with data enrichment is 11% to 27% lower than that of XGBoost and LSTM without data enrichment. In the future, variations on the data enrichment methods can be further explored as a promising direction of enhancing short-term wind power forecasting performance. [ABSTRACT FROM AUTHOR]

The transport sector is one of the main contributors to gas emissions and the greenhouse effect. Mobility as a Service seeks to combine different means of transport (public transport, bicycles, cars, etc.) in an integrated way to solve this problem. Raising drivers' awareness and advising them to drive more efficiently and economically can help reduce environmental impact. Ultimately, eco-driving can help save fuel, money and provide higher passenger satisfaction during the trip. The research uses a contextually enriched automotive data set collected via OBD, during passenger transportation in private cars. A multilevel analysis was conducted using a variety of approaches. With AI-based analysis, using unsupervised learning, we identified five groups of similarities as a function of speed and acceleration. The efficiency of each driver was examined and speed and acceleration/deceleration were noted as the most influential factors. Also, passengers' perception of driving quality is negatively affected by sudden acceleration and deceleration. Passengers have a poor perception of the driving efficiency and additional efforts should be made to educate them about the environmental aspect of driving. Finally, the main contribution is the eco-aware framework, which covers and describes the entire process from data collection to analysis. The collected data are publicly available and include 110 trips represented by more than 65,000 data points. [ABSTRACT FROM AUTHOR]

Recently, there has been an increasing interest in extracting and annotating tables on the Web. This activity allows the transformation of textual data into machine-readable formats to enable the execution of various artificial intelligence tasks, e. g., semantic search and dataset extension. Semantic Table Interpretation (STI) is the process of annotating elements in a table. The paper explores Semantic Table Interpretation, addressing the challenges of Entity Retrieval and Entity Disambiguation in the context of Knowledge Graphs (KGs). It introduces LamAPI , an Information Retrieval system with string/type-based filtering and s-elBat , an Entity Disambiguation technique that combines heuristic and ML-based approaches. By applying the acquired know-how in the field and extracting algorithms, techniques and components from our previous STI approaches and the state of the art, we have created a new platform capable of annotating any tabular data, ensuring a high level of quality. [ABSTRACT FROM AUTHOR]

Diminishing viability of manual fault diagnosis in the increasingly complex emerging cellular network has motivated research towards artificial intelligence (AI)-based fault diagnosis using the minimization of drive test (MDT) reports. However, existing AI solutions in the literature remain limited to either diagnosis of faults in a single base station only or the diagnosis of a single fault in a multiple BS scenario. Moreover, lack of robustness to MDT reports spatial sparsity renders these solutions unsuitable for practical deployment. To address this problem, in this paper we present a novel framework named Hybrid Deep Learning-based Root Cause Analysis (HYDRA) that uses a hybrid of convolutional neural networks, extreme gradient boosting, and the MDT data enrichment techniques to diagnose multiple faults in a multiple base station network. Performance evaluation under realistic and extreme settings shows that HYDRA yields an accuracy of 93% and compared to the state-of-the-art fault diagnosis solutions, HYDRA is far more robust to MDT report sparsity.

The ubiquitous availability of datasets has spurred the utilization of Artificial Intelligence methods and models to extract valuable insights, unearth hidden patterns, and predict future trends. However, the current process of data collection and linking heavily relies on expert knowledge and domain-specific understanding, which engenders substantial costs in terms of both time and financial resources. Therefore, streamlining the data acquisition, harmonization, and enrichment procedures to deliver high-fidelity datasets readily usable for analytics is paramount. This paper explores the capabilities of SemTUI , a comprehensive framework designed to support the enrichment of tabular data by leveraging semantics and user interaction. Utilizing SemTUI, an iterative and interactive approach is proposed to enhance the flexibility, usability and efficiency of geospatial data enrichment. The approach is evaluated through a pilot case study focused on urban planning, with a particular emphasis on geocoding. Using a real-world scenario involving the analysis of kindergarten accessibility within walking distance, the study demonstrates the proficiency of SemTUI in generating precise and semantically enriched location data. The incorporation of human feedback in the enrichment process successfully enhances the quality of the resulting dataset, highlighting SemTUI's potential for broader applications in geospatial analysis and its usability for users with limited expertise in manipulating geospatial data. [ABSTRACT FROM AUTHOR]

The shift towards renewable energy integration into smart grids has led to complex management processes, which require finer-grained energy and heat generation/ demand forecasting while considering data from monitoring devices and the integration of smaller multi-energy sub-systems at the community, district, or buildings level. However, energy prediction is challenging due to the high variability in the electrical and thermal energy demands of building occupants, the heterogenous characteristics of the energy assets or buildings in a district, and the length of the forecasting horizon. In this paper, we define a data-driven machine-learning pipeline to predict the electricity and thermal consumption of buildings and energy assets from a city district in 24 h intervals. Each pipeline’s step is divided into sensors’ data processing and model integration, data enrichment and features engineering, and multilayer perceptron model training. To address some of the drawbacks of using the multi-layer perceptron model, such as slow convergence rate and risk of overfitting, and to ensure a lower error in the energy prediction process features, an engineering technique was employed. We incorporated weather data features and interaction features derived from fusing the energy data with statistical models to capture the nonlinear patterns of the electrical and heat demands. The proposed approach was successfully validated in a real-world environment, a city district in Gent, Belgium. It featured good prediction results for electricity and heat production and consumption of various assets without considering the physical characteristics, making it viable and easily applicable in broader urban areas. The evaluation of energy prediction accuracy yielded good results, with a Mean Absolute Error (MAE) falling within the range of 0.003 to 3.27, and a Mean Absolute Scaled Error (MASE) ranging from 7 × 10−5 to 2.57 × 10−3.