Λογότυπο αποθετηρίου

Πλοήγηση ανά Συγγραφέα "Maggina, Spiridoula"

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
Α Β Γ Δ Ε Ζ Η Θ Ι Κ Λ Μ Ν Ξ Ο Π Ρ Σ Τ Υ Φ Χ Ψ Ω
Φιλτράρετε τα αποτελέσματα πληκτρολογώντας τα πρώτα γράμματα
Τώρα δείχνει 1 - 1 από 1
  • Μικρογραφία εικόνας
    Τεκμήριο
    An analysis of the Greek 1991-2016 crime data
    (2018-11-28) Maggina, Spiridoula; Μαγγίνα, Σπυριδούλα; Athens University of Economics and Business, Department of Statistics; Demiris, Nikolaos
    The study of crimes has great scientific interest with practical implications regarding government future strategies aiming at society safety as well as and other social and macroeconomic impact.In this study, we have utilized annual historical crime data for the period 1991-2016, concerning seven common crimes in Greece's territory general,as well as focusing on the main 14 districts that form the country.The stationary time series is one of the tools for making predictions and autoregressive integrated moving average (ARIMA) models have been already successfully used in forecasting econometrics and other social science problems.The main goal of this study was the prediction of the best fitted models for depicting the crime trend and forecasting future values,regarding the whole country and Greece's 14 individual regions. In this thesis, we have used Box Jenkins ARIMA methodology for the univariate time series analysis,after we removed the trend from the series.We suggest, for the time series analysis of count data,the implementation of a method based on GLM models,performed from the R package "tscount".We examined also, a multivariate analysis with VAR models.The applied methods explain the time series adequately. According to the forecasts we do not expect significant changes in the crime pattern in the future. Regarding the variable “fraud” however, a substancial increase is been expected for the next 30 years in total Greece and some specific regions namely An.Makedonia-Thraki, Thessalia, Ipeiros, Peloponnisos, Attiki, Ionia Nisia, Voreio and Notio Aigaio according to the predictive models. ARIMA models could perform better in shorter run forecasts and with no doubt longer time series could provide better results.