The selection of optimal subset of features from high-dimensional datasets still remains a major challenge during breast cancer detection and categorization. There exist several research works regarding optimal featu...
详细信息
The selection of optimal subset of features from high-dimensional datasets still remains a major challenge during breast cancer detection and categorization. There exist several research works regarding optimal feature subset selection from high-dimensional datasets, but the obtained results are not satisfying when multidimensional data sets (MDDs) are employed in large amount during disease analysis. In this article, an effective feature subset selection and classification method suitable for MDD is proposed. At first, the important and distinct features are extracted from the mammogram images using a Deep Neural Network with wrapper-based extraction technique. Then, a novel two-phase mutation strategy integrated with grey wolf optimizer algorithm is employed for selecting the most relevant feature subsets. Finally, a learning-based semilazy Bayesian network classifier with parallel implementation is proposed for the precise categorization of the breast cancer stages. The proposed method is executed in MATLAB platform and analyzed using mammogram images taken from MAMMOsetdatabase. The proposed method is likened with three state-of-the-art existing feature subset selection and classification approaches for validating the efficiency of the proposed approach. For dataset 1, the proposed method shows an accuracy of 90%, 92% and 98%, which is better than the existing methods taken for comparison. Also for dataset 2, the proposed method shows an accuracy of 97%, 91% and 94%, which is much better than the accuracy achieved by the existing approaches. Thus, the proposed approach outperforms the compared existing approaches by providing better precision, recall, F-measure, specificity and accuracy.
Two-dimensional (2D) materials exhibit unique optical properties when controlled to atomic thickness, and show large potential for applications in optoelectronics, photodetectors, and tunable excitonic devices. Curren...
详细信息
Two-dimensional (2D) materials exhibit unique optical properties when controlled to atomic thickness, and show large potential for applications in optoelectronics, photodetectors, and tunable excitonic devices. Current characterization techniques, including conventional optical microscopy, atomic force microscopy (AFM), and Raman spectroscopy are time-consuming and labor-intensive for studying large-scale samples. To realize the rapid identification of monolayer and few-layer crystals in the "haystack" of hundreds of flakes appearing in the exfoliation process, line-scan hyperspectral imaging microscopy combined with linear unmixing was developed to identify 2D molybdenum disulfide (MoS2) and hexagonal boron nitride (hBN) samples. A complete hyperspectral measurement and analysis, including single-band analysis, pixel-level spectral analysis and image classification was performed on MoS2 and hBN flakes with mono- and few-layer thickness. The characteristic spectra were extracted and analyzed via linear unmixing calculations to reconstruct the distribution images. The abundance maps showed the spatial distribution of these flakes with flake positions output, realizing an automatic identification of target flakes. This work shows a rapid and robust method for the determination of abundance maps of 2D flakes distributed over macroscopic areas.
The unique optical properties of two-dimensional (2D) materials are largely dependent on the number of atomic layers. Hyperspectral imaging microscopy shows large potential for rapid and accurate thickness mapping. To...
详细信息
The unique optical properties of two-dimensional (2D) materials are largely dependent on the number of atomic layers. Hyperspectral imaging microscopy shows large potential for rapid and accurate thickness mapping. To process the acquired hyperspectral dataset and to deal with pixel-level spectra remain a challenge for further application. In this work, two quantitative classification strategies including linear unmixing and spectral peak mapping were conducted to characterize a multilayer semiconducting MoS2 flake with nanoscale thickness variations. The comparative study paves the way to identify 2D semiconducting materials with random layer numbers (monolayer, bilayer, and few-layer) in both laboratory and industry.
Hive is a data warehouse architecture in cloud computing. In order to solve the inadequate of massive data storage, query, and computing power in current electric power data warehouse, a method of electric power data ...
详细信息
ISBN:
(纸本)9783037855300
Hive is a data warehouse architecture in cloud computing. In order to solve the inadequate of massive data storage, query, and computing power in current electric power data warehouse, a method of electric power data warehouse based on Hive is proposed. Combining data analysis demands of electric power entreprises, the integration architecture between Hive and column-oriented storage is designed in electric power data warehouse, and the process of which is also given. At last, electric power equipment condition data is used for experiment on Hadoop cluster, results show that this method can quickly achieve query and analysis in massive multidimensional data set.
Using a database management system ( DBMS) is essential to ensure the data integrity and reliability of large, multidimensional data sets. However, loading multiterabyte data into a DBMS is a time-consuming and error-...
详细信息
Using a database management system ( DBMS) is essential to ensure the data integrity and reliability of large, multidimensional data sets. However, loading multiterabyte data into a DBMS is a time-consuming and error-prone task that the authors have tried to automate by developing the sqlLoader pipeline - a distributed workflow system for data loading.
A set of principal axes accounts for the largest variation directions of a dataset. It is one of the important general features of the dataset. The paper discusses the forecast modelling method for the rotation of p...
详细信息
A set of principal axes accounts for the largest variation directions of a dataset. It is one of the important general features of the dataset. The paper discusses the forecast modelling method for the rotation of principal axes in the multidimensional space. The method was used in forecasting and analysing the development of city economy in China. The movement direction of the main features of the economic development of the city group can be predicted. (C) 1998 :Elsevier Science B.V. All rights reserved.
暂无评论