China 
Africa 
Explore chapters and articles related to this topic
Machine Learning
Published in Pedro Larrañaga, David Atienza, Javier Diaz-Rozo, Alberto Ogbechie, Carlos Puerto-Santana, Concha Bielza, Industrial Applications of Machine Learning, 2019
Pedro Larrañaga, David Atienza, Javier Diaz-Rozo, Alberto Ogbechie, Carlos Puerto-Santana, Concha Bielza
Affinity propagation algorithm (Frey and Dueck, 2007) is based on the concept of message passing between objects. The goal is to find a subset of cluster prototypes, referred here as exemplars, that are members of the input dataset. All objects are simultaneously considered as potential exemplars, avoiding the selection of initial cluster prototypes. The number of clusters, K, does not have to be determined before running the algorithm.
Client profile prediction using convolutional neural networks for efficient recommendation systems in the context of smart factories
Published in Enterprise Information Systems, 2022
Nadia Nedjah, Victor Ribeiro Azevedo, Luiza De Macedo Mourelle
In (Chatzilari et al. 2012), the selection of a set of images of the social network Flickr is made to detect objects. Clustering of image regions using the Affinity Propagation algorithm is done. The authors use supervised learning in object classification with image recognition models. The authors use of the Probabilistic Latent Semantic Analysis (PLSA) algorithm and the Dirichlet Latent Allocation (LDA) algorithm for probability distribution of the characteristics of images. They create visual vocabulary using the K-Means algorithm, which is a clustering method to partition n observations among k groups, where each observation belongs to the group closest to the mean. A binary classifier was also used with the supervised support vector machine technique.
Revealing representative day-types in transport networks using traffic data clustering
Published in Journal of Intelligent Transportation Systems, 2023
Matej Cebecauer, Erik Jenelius, David Gundlegård, Wilco Burghout
The set of representative day-types can be determined using varying clustering methods, similarity or distance measures among days, and with a varying number of clusters (Estivill-Castro, 2002). Each cluster would then represent one representative network-wide day-type. Many different cluster validity indices or evaluation metrics could be used for selecting the clustering which best matches the criteria. Common practice is to use one of the well-known and widely accepted clustering methods such as k-means, hierarchical, affinity propagation, or spectral clustering (Lopez et al., 2017; Yang et al., 2017; Krishnakumari et al., 2020; Ferranti, 2020; Cebecauer et al., 2019; Chiabaut & Faitout, 2021).
Detecting events from the social media through exemplar-enhanced supervised learning
Published in International Journal of Digital Earth, 2019
Xuan Shi, Bowei Xue, Ming-Hsiang Tsou, Xinyue Ye, Brian Spitzberg, Jean Mark Gawron, Heather Corliss, Jay Lee, Ruoming Jin
It can be concluded and expected that in the future, classifiers built upon all datasets 1 to 5 will be able to clarify whether a new tweet message is really relevant to wildfire events or not with high confidence. Eventually, parallel and high-performance computing solutions may be required when big data are processed, particularly because the Affinity Propagation calculation produces a significant bottleneck in handling the source data and intermediate data. The required memory of such processes increase dramatically beyond the normal computational capacity of a desktop computer, while classification by SVM over big data is another time-consuming task that has to be processed efficiently.