Prediction and interpretation of events through variable relevance analysis in machine learning models
Regression and Classification techniques has been studied in Machine learning in order to solve several tasks. In particular, the creation of models based both on predicting and interpretability is one of the central themes of this work. The development of this literature has been possible, due to t...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Fecha de publicación: | 2022 |
| País: | España |
| Institución: | Universidad de Huelva (UHU) |
| Repositorio: | Arias Montano. Repositorio Institucional de la Universidad de Huelva |
| Idioma: | inglés |
| OAI Identifier: | oai:ariasmontano.uhu.es:10272/27527 |
| Acceso en línea: | https://hdl.handle.net/10272/27527 |
| Access Level: | acceso abierto |
| Palabra clave: | Regression Classification Machine learning Regresión Clasificación 5302.02 Modelos Econométricos 5302.05 Series Cronológicas Económicas 1209.15 Series Temporales 1209.14 Técnicas de Predicción Estadística |
| Sumario: | Regression and Classification techniques has been studied in Machine learning in order to solve several tasks. In particular, the creation of models based both on predicting and interpretability is one of the central themes of this work. The development of this literature has been possible, due to the necessity of both predicting and describing the reality learnt by statistical models. In this regard, Most representative decision-tree ensemble methods has been used not only for predict events but also to examine the variable importance in order to understand the elements that inform and make it relevant to understand diverse phenomena. However, the literature and the empirical applications are still scarce. For this reason, this the-sis tries to empirically analyze these models and to develop new models that allow progress in the understanding of the relationships and relevance between variables in the field of econometrics. Chapter 2 proposes a new accurate model for US eco-nomic recessions giving as an output of the work the most important treasury term spreads and rules for US economic recession detection, finding the most relevant term spread found is 3-month–6-month, which is proposed to be monitoring by eco-nomic authorities. Chapter 3 proposes a new accurate model in order to study the relevance of price, GDP and affordability as a mechanism for controlling the demand for cigarettes, finding that although the demand functions estimated so far are useful to make predictions about the behavior of cigarette demand, the government must consider that price is a good tool to control tobacco consumption from a certain point of affordability. Finally, Chapter 4 analyze if the EPSs established in Spanish provinces were fulfilled and the anomalies observed in provinces where sales exceed expected values are measured, finding that the provinces in which sales below rea-sonable values are observed (as detected by the EPSs) present a clear geographical pattern. Furthermore, the values provided by the EPSs in Spain, as indicated in the previous literature, are slightly oversized. Finally, there are regions bordering other countries or with a high tourist influence in which the observed sales are higher than the expected values. ----------------------------------------------------------------------------------------- |
|---|