| title | author | geometry | output |
|---|---|---|---|
Exercise 3 -- Work Plan |
Backé Julian, Russo Dominik, Salzer Tobias, Singh Ajayvir |
left=18mm,right=18mm,top=6mm,bottom=21mm |
pdf_document |
\vspace{-13mm}
\vspace{-3mm}
-
How do university rankings change over time?
The goal is to compare and visualize rank(s) over time. It will be interesting to find out the methodology and the criteria of the rank calculation (focus on research or on teaching).
-
Which characteristics of universities contribute most to good rankings, or to large changes in the ranking position?
Analyse individual criteria in datasets and what characteristics contribute most to good ranking. Following attributes could be interesting for this question:
\begin{minipage}[t]{0.45\textwidth} \begin{itemize} \item Budget for conducting research \item Tuition fee \item Presence of faculties in a certain field (research focus of the university) \item Age of university \item Sector (public/private) \item Research output (published papers per annum)
\end{itemize} \end{minipage} \hspace{5mm} \begin{minipage}[t]{0.45\textwidth} \begin{itemize}
\item Eesearch impact (citations per paper) \item Male to female ratio (students, staff,...) \item Number of students (total or in certain fields) \item Number of notable alumni to total number of\\ students ratio \item Students to resource ratio (labs, computers, ...) \item Students to professors ratio\end{itemize} \end{minipage}
For this, an extensive data-exploration and effective visualisations will be necessary.
-
How do these characteristics correlate with characteristics of cities or countries in which the university is located?
Following location related attributes could be considered: \begin{minipage}[t]{0.45\textwidth} \begin{itemize} \item GDP \item HDI \item Corruptions perception index per country \item City size (number of inhabitants) \item Average ranking grouped by continent, country or city over time
\end{itemize} \end{minipage} \hspace{5mm} \begin{minipage}[t]{0.45\textwidth} \begin{itemize}
\item Public expenditure on education \item Rain days per year \item Absolute number of universities per continent, country or city \item Relative (per capita) number of universities per continent, country or city\end{itemize} \end{minipage}
-
Are there predictors for increases or decreases in the rankings?
The above attributes should be analysed and prediction models for the rank trend of a university should be built.
\vspace{-2mm}
\vspace{-3mm} \begin{minipage}[t]{0.45\textwidth} \begin{itemize}
\item Corruption in public sector: (\hyperlink{https://data.worldbank.org/indicator/IQ.CPA.TRAN.XQ}{click here})
\item GDP Growth: (\hyperlink{https://data.worldbank.org/indicator/NY.GDP.MKTP.KD.ZG}{click here})
\item HDI: (\hyperlink{https://databank.worldbank.org/reports.aspx?source=1147\&series=UNDP.HDI.XD}{click here})
\item Top 2000 universities 2019/2020: (\hyperlink{https://cwur.org/2019-2020.php}{click here})
\end{itemize} \end{minipage} \hspace{5mm} \begin{minipage}[t]{0.45\textwidth} \begin{itemize}
\item University rankings: (\hyperlink{https://www.kaggle.com/mylesoneill/world-university-rankings}{click here})
\item Weather data: (\hyperlink{https://earthdata.nasa.gov/}{click here})
\item World population metrics: (\hyperlink{https://sedac.ciesin.columbia.edu/data/collection/gpw-v3>}{click here})
\end{itemize} \end{minipage}
\vspace{-2mm}
\vspace{-3mm} Step 1: Get the data. Some further data source research has to be done. The data will be loaded into a Jupyter Notebook (in Pandas Dataframes) and cleaned up if necessary. Each group member will have his own information to search for and his own datasets to clean and load in.
Step 2: Explore the data. As already noted, the data exploration represents an essential part of this work. This includes in particular finding out about (cor)relations between variables and the ranking by calculating correlation coefficients as well as plotting the data in an appropriate manner. Each group member will explore his datasets, which will be linked to the dataset containing the ranking.
Step 3: Model the data. Once possible predictors for increases or decreases of rankings were identified, we would like to find out how the predictors interact with the rankings. For this, we aim to to build up simple machine learning models in the beginning. In a next step, these could be improved by more complex approaches identified during the work progress. This work will be split up when the outcomes of the data exploration are manifest.