Online Anomaly detection in operational data from dynamic systems is challenging due to the inherent complexity and variability of these datasets. This research introduces an innovative approach that leverages the Sli...
详细信息
Many telecommunication companies try to predict customer churn used supervised learning. This article studies the critical condition in the telecommunications services industry (telco) by using analytics tools of unsu...
详细信息
ISBN:
(纸本)9781665425803
Many telecommunication companies try to predict customer churn used supervised learning. This article studies the critical condition in the telecommunications services industry (telco) by using analytics tools of unsupervised learning. We examine seven different unsupervised algorithms for solving the most crucial assets for a business in numerous dynamic and competitive telecommunication companies within a marketplace, which the data available in Kaggle. The results indicate that the use of unsupervised algorithms led to keep the customers are most likely to churn. Based on our unsupervised results, some suggestions for improving customer churn prediction by supervised learning are also made.
During oil well drilling, geological knowledge of the layers to be drilled is essential to dimension the drilling parameters and plan contingency operations in case of operational failures. The Ariri Formation, lying ...
详细信息
During oil well drilling, geological knowledge of the layers to be drilled is essential to dimension the drilling parameters and plan contingency operations in case of operational failures. The Ariri Formation, lying in the Santos Basin (southeastern Brazil), can reach a thickness exceeding 3000 m and consists of salts with a complex rheology, requiring some effort to understand these minerals' vertical and lateral distributions. This study aims to classify the electrofacies of evaporitic sequences in a semi-automated way based on unsupervised algorithms applied to geophysical well-logs and drilling parameters. The results were validated based on the previous classifications performed by interpreters, in which six types of saline minerals predominate in the wells under study: halite, anhydrite, tachyhydrite, carnallite, sylvite, and sylvinite. The database consisted of a set of fourteen wells located in the offshore portion of the Santos Basin. unsupervised analyses are developed using the multilayer perceptron with lateral connections (MPLC), k-means, and Self-Organising Maps (SOM) algorithms. The obtained clusters are classified according to their mineralogical composition and drilling resistance. The SOM and MPLC algorithms provide the best accuracy in segmenting the main evaporite groups and highlighting possible mixtures between them. In terms of facies, the clustering provides electrofacies with different levels of drilling resistance for the same mineral. The selection of the best grouping enables a detailed subdivision for the Ariri Formation, which will serve as a basis for future stratigraphic studies in the distal setting of the Santos Basin.
The identification of digital market segments to make value-creating propositions is a major challenge for entrepreneurs and marketing managers. New technologies and the Internet have made it possible to collect huge ...
详细信息
The identification of digital market segments to make value-creating propositions is a major challenge for entrepreneurs and marketing managers. New technologies and the Internet have made it possible to collect huge volumes of data that are difficult to analyse using traditional techniques. The purpose of this research is to address this challenge by proposing the use of AI algorithms to cluster customers. Specifically, the proposal is to compare the suitability of supervised algorithms, XGBoost, versus unsupervised algorithms, K-means, for segmenting the digital market. To do so, both algorithms have been applied to a sample of 5 million Spanish users captured between 2010 and 2022 by a lead generation start-up. The results show that supervised learning with this type of data is more useful for segmenting markets than unsupervised learning, as it provides solutions that are better suited to entrepreneurs' commercial objectives.
The rapid development of the Internet not only provides convenience but also causes a variety of abnormal information. In general, the influence of abnormal information on different fields is far more important than t...
详细信息
Nanoparticulate electrocatalysts for the oxygen reduction reaction are structurally diverse materials. Scanning transmission electron microscopy (STEM) has long been the go-to tool to obtain high-quality information a...
详细信息
Nanoparticulate electrocatalysts for the oxygen reduction reaction are structurally diverse materials. Scanning transmission electron microscopy (STEM) has long been the go-to tool to obtain high-quality information about their nanoscale structure. More recently, its four-dimensional modality has emerged as a tool for a comprehensive crystal structure analysis using large data sets of diffraction patterns. In this study, we track the alternations of the crystal structure of individual carbon-supported PtCu3 nanoparticles before and after fuel cell-relevant activation treatment, consisting of a mild acid-washing protocol and potential cycling, essential for forming an active catalyst. To take full advantage of the rich, identical location 4D-STEM capabilities, unsupervised algorithms were used for the complex data analysis, starting with k-means clustering followed by non-negative matrix factorization, to find commonly occurring signals within specific nanoparticle data. The study revealed domains with (partially) ordered alloy structures, twin boundaries, and local amorphization. After activation, specific nanoparticle surface sites exhibited a loss of crystallinity which can be correlated to the simultaneous local scarcity of the ordered alloy phase, confirming the enhanced stability of the ordered alloy during potential cycling activation conditions. With the capabilities of our in-house developed identical-location 4D-STEM approach to track changes in individual nanoparticles, combined with advanced data analysis, we determine how activation treatment affects the electrocatalysts' local crystal structure. Such an approach provides considerably richer insights and is much more sensitive to minor changes than traditional STEM imaging. This workflow requires little manual input, has a reasonable computational complexity, and is transferrable to other functional nanomaterials.
Most of the distributed photovoltaics (PV) are installed behind the meter (BTM), single-meter deployments permit distribution system operators to monitor only the net load and exclude the BTM PV generation, so the gro...
详细信息
Most of the distributed photovoltaics (PV) are installed behind the meter (BTM), single-meter deployments permit distribution system operators to monitor only the net load and exclude the BTM PV generation, so the growing prevalence of BTM PV installations negatively affects distribution system planning and the local balance of supply and demand. However, existing methods for net load disaggregation mainly rely on the installation of expensive monitoring devices and high-resolution sensors, and face challenges such as privacy concerns, data diversity, and communication barriers. In this paper, an unsupervised method for aggregated net load disaggregation is proposed that achieves accurate separation of BTM PV outputs and actual loads using only net load data and exogenous variables. First, a data-driven method is developed to construct the actual load sample matrix of customers. Then, a virtual PV sample construction method based on the self-feedback decoupling algorithm (SFDA) is proposed to tackle the invisibility of BTM PV resources. The method performs self-feedback learning and constructs the virtual PV samples by minimizing the long-term decomposition residuals, and generates the virtual PV sample matrix. Finally, the model learning results are employed to achieve net load disaggregation through the contextually supervised source separation (CSSS) algorithm. The study utilized real open-source data whereby analyses reveal the method greatly enhances the decoupling accuracy of unsupervised algorithms. Furthermore, it eliminates a series of problems associated with traditional supervised algorithms and expands the scope of unsupervised decoupling methods.
The bridge network is progressively aging, with an alarming proportion of bridges over 100 years. This situation engenders substantial risks to the overall reliability of transportation networks, requiring innovative ...
详细信息
The bridge network is progressively aging, with an alarming proportion of bridges over 100 years. This situation engenders substantial risks to the overall reliability of transportation networks, requiring innovative methods for efficient management. Monitoring can provide a direct source of information about structural behavior generating alerts when changes occur. Real-time alerts enable effective infrastructure management and decisionmaking during damage or anomalous situations. However, monitoring can result in a large amount of data that is often difficult to convert into valuable information in real time. This paper presents an approach for realtime detection of abrupt damage occurrence in bridges using unsupervised anomaly detection algorithms and strain/acceleration measurements. The approach incorporates the separation of measurements into events having the same loading nature and the construction of three feature matrices based on statistical features, timefrequency features, and wavelet spectrum features. It includes the evaluation of five anomaly detection algorithms including Isolation Forest, One-Class Support Vector Machine, Robust Random Cut Forest, Local Outlier Factor, and Mahalanobis Distance. The approach is illustrated with a case study of a steel-bascule-railway bridge, that has experienced a brittle cracking event during monitoring. Results highlight the robustness of One-Class Support Vector Machine, Isolation Forest, and Local Outlier Factor algorithms in promptly detecting abrupt changes across different features. The separation of strain and acceleration data into loading-based events, coupled with the comparison of previous and new event features, provides robust feature matrices for effective damage detection. Enhanced detection and higher scores are particularly attributed to time-frequency domain features during damage occurrence. The presented approach can be used as a base on how to perform real-time anomaly detection within the cont
User behaviour, human mistakes, and underperforming equipment contribute to wasted energy in buildings and industries. Identifying anomalous consumption power behaviour can help to reduce peak energy usage and change ...
详细信息
User behaviour, human mistakes, and underperforming equipment contribute to wasted energy in buildings and industries. Identifying anomalous consumption power behaviour can help to reduce peak energy usage and change undesirable user behaviour. Furthermore, decreasing energy consumption in buildings is difficult because usage patterns vary from one building to the next. So, the main contribution in this manuscript is to propose a lightweight architecture for smart meter to identify abnormalities in power consumption for each building individually using machine learning (ML) models and implement on a Single Board Computer. To detect daily and periodic pattern anomalies, two models of anomaly detection based on supervised and unsupervised ML algorithms are built and trained where numerous algorithms were utilised to select the best algorithm for each model. Also, the proposed approach enables iterative procedure modifications by retraining the two anomaly detection models on data aggregator server based on the received data meter from the specific smart meter to give better power service to clients while minimising provider losses. The effectiveness and efficiency of the suggested approach have been proven through extensive analysis.
Predictive maintenance is a crucial strategy in smart industries and plays an important role in small and medium-sized enterprises (SMEs) to reduce the unexpected breakdown. Machine failures are due to unexpected even...
详细信息
ISBN:
(纸本)9783031143434;9783031143427
Predictive maintenance is a crucial strategy in smart industries and plays an important role in small and medium-sized enterprises (SMEs) to reduce the unexpected breakdown. Machine failures are due to unexpected events or anomalies in the system. Different anomaly detection methods are available in the literature for the shop floor. However, the current research lacks SME-specific results with respect to comparison between and investment in different available predictive maintenance (PdM) techniques. This applies specifically to the task of anomaly detection, which is the crucial first step in the PdM workflow. In this paper, we compared and analyzed multiple anomaly detection methods for predictive maintenance in the SME domain. The main focus of the current study is to provide an overview of different unsupervised anomaly detection algorithms which will enable researchers and developers to select appropriate algorithms for SME solutions. Different Anomaly detection algorithms are applied to a data set to compare the performance of each algorithm. Currently, the study is limited to unsupervised algorithms due to limited resources and data availability. Multiple metrics are applied to evaluate these algorithms. The experimental results show that Local Outlier Factor and One-Class SVM performed better than the rest of the algorithms.
暂无评论