An in-depth review of SOMs with applications

e-Archivo Repository

Show simple item record

dc.contributor.advisor García Portugues, Eduardo Jiménez Rama, Óscar Manuel 2020-05-28T10:50:59Z 2020-05-28T10:50:59Z 2019-09 2019-10-16
dc.description.abstract The development of internet has opened a new gate from where data can be collected. It is expected that in 2020, each human being will create 1,7 MB of information per second. The management of this huge quantity of data brings new problems and limitations. New analysis techniques are needed to overcome such obstacles to ultimately extract actionable insights from high volumes of data. The adequate analysis of this massive information, not only can be used in the business world to solve problems and take optimal decisions, it also pushes forward research and innovation in many fields of study. From a statistical point of view, more information not always leads to better predictions: when data “grows”, this usually implies the analysis of higher dimensional spaces in which more complex data structures may be present. Dimensionality reduction techniques, such as Principal Components Analysis (PCA), were conceived to give solution to this problem. Many of these techniques are of a linear nature and, as a consequence, hold limitations when applied to complex data structures. For this reason, many non linear procedures have been proposed in the last decades. Self-Organizing Maps (SOMs) have been gaining popularity over the years with the rise of computing power and since they bring a distinctive approach for high dimensional data inspection focusing on data clustering. SOMs ability to map high dimensional data into a low-dimensional space (commonly two-dimensional) is motivated by certain evidences found in cerebral nature. More specifically, with the studies regarding auto associative memory and its relation with neurons ability to learn. In this Bachelor Thesis, a complete review of SOMs is carried out. First, theoretical foundations are addressed, contextualizing the technique in its corresponding category within machine learning. Second, an in-depth description from a mathematical point of view is presented with a pedagogic aim. Third, software implementation is discussed using practical examples applied to synthetic data, in order to illustrate how the algorithm works in practice. Finally, an application of SOMs, related with the performance of football players, is presented in order to exploit its properties and extract insights from data. The codes of the examples can be openly accessed from the following GitHub repository:
dc.language.iso eng
dc.rights Atribución-NoComercial-SinDerivadas 3.0 España
dc.subject.other Self-Organizing Maps
dc.subject.other Data Science
dc.subject.other Unsupervised Learning
dc.subject.other Dimensionality Reduction
dc.title An in-depth review of SOMs with applications
dc.type bachelorThesis
dc.subject.eciencia Estadística
dc.subject.eciencia Telecomunicaciones
dc.rights.accessRights openAccess Ingeniería en Tecnologías de Telecomunicación
dc.contributor.departamento Universidad Carlos III de Madrid. Departamento de Estadística
 Find Full text

Files in this item

*Click on file's image for preview. (Embargoed files's preview is not supported)

The following license files are associated with this item:

This item appears in the following Collection(s)

Show simple item record