Statistical learning: Bernoulli time series modelling for discrete decision choice
Informar que el próximo 24-26 de junio 2021 presentaré mi investigación en Statistical Learning en el "4th Internacional Conference on Econometrics and Statistics (EcoSta 2021)" en Hong Kong.
Title: Statistical learning: Bernoulli time series modelling for discrete decision choice
Authors: Miguel Angel Ruiz Reina - Universidad de Malaga (Spain)
Keywords: decision making, forecasting, generalized methods of moments, statistical learning
Abstract: Not infrequently, decisions have been temporally binary modelled under uncertainty. Binary data time series are used in natural science, pure science, computer science, or social science. We present the temporal Bernoulli modelling in uncertainty contexts that allows extracting knowledge for Statistical Learning, finding patterns not previously described and without information. The proposed Bernoulli regression uses an uncertainty factor with high explanatory power; firstly, we can find endogeneity problems with the error term. The Generalised Method of Moments method, solving endogeneity problems and guaranteeing the consistency of the parameters. Our goal is to intelligently provide future decision tools with modelling the past in future uncertainty contexts. Our model is compared with other prediction models in the literature (Entropy Model, SARIMA and ARDL + Seasonality). The Matrix U1 Theil is a decision tool that guarantees the model's high forecasting capacity presented over the others. The decision-making framework can be applied in a wide variety of domains. This modelling can be applied to supervised learning and optimisation contexts. This article concludes with the analysis of real data sets that confirm the methodological framework; in particular, we are interested in modelling economic agents' decision temporarily in a mutually exclusive context.
Miguel Ángel Ruiz Reina