Πλοήγηση ανά Συγγραφέα "Tsaka, Matilnta"

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

Τώρα δείχνει 1 - 1 από 1

Demand forecasting for the bike sharing system in Washington DC
(2021-07-21) Tsaka, Matilnta; Τσάκα, Ματίλντα; Athens University of Economics and Business, Department of Management Science and Technology; Karlis, Dimitrios
The aim of this project is to study the demand of bike sharing systems and try to understand what other variables influence the total number of rentals. The main objective is to create a predictive model for the demand of 10 stations of the Capital Bikeshare, one of the U.S.A.’s largest bicycle sharing systems on an hourly basis. At first, I did some research on the topic and the methodologies that have been used to develop a predictive model for the system’s demand forecasting. The next step was to acquire the necessary data for our analysis. This required downloading data from Capital Bikeshare website, Washington D.C.’s local government website (https://www.capitalbikeshare.com/system-data) and weather data from an internet weather service website (www.timeanddate.com.). In this project we are going to use historical data for two years from 2018 to 2019 on an hourly basis. The data mentioned above were aggregated into one single dataset. For our analysis, we chose 10 stations of the bike sharing system to study. Once the data for the 10 selected stations were gathered, the R programming language was used to visualize and explore the data. Then, time series forecasting models were built to predict the demand of bikes of each station. For most of the station the models gave us good predictions, but they were not the best that we could get. After fitting the models, it was obvious that data incorporated more than one seasonal component, and this was something that the models that were trained could not handle to result in better prediction.