Screen scraping Alumni directory with Python in order to determine the average duration to finish each degree.

I have been always curious about how much time does it take to become an engineer in Spain. I have always heard that engineering degrees are extremely difficult and for that reason people usually need more time than the official, but I never found a liable source for that information.

I have studied in Universidad Carlos III de Madrid, and been an Alumni allows me to search through the Alumni directory (if you are an alumni you can access that directory here). In that place I could found any student that has finished the studies in this university, giving information about its name, the degree coursed and the starting and finishing year. So here I have a good source of information to investigate.

Sometime ago I experimented a bit with BeautifulSoup library to parse the HTML code so I thought this could be a good project to apply that in a real context.

Note that this is a personal project that I've been developing while learning about Python and some of its libraries, so don't expect to find an astonishing quality code, because you won't ;-)

In order to retrieve all the information from that directory I have created a Python script called download_info_alumni_uc3m.py that basically could call one of two basic functions implemented on alumni.py:

• all() Retrieve all the alumni data from all the degrees and save it on /data/YYYY-MM-DD/alumni/ creating different csv files, one for each degree. It also creates a degrees.csv file on /data/YYYY-MM-DD/ that relates the degree name with the associated code used in the previous degree-related csv files (like 19;T for Ingenieria en Telecomunicacion).
• engineering() Retrieve alumni data from all the engineering degrees and save it on /data/YYYY-MM-DD/alumni/ creating different csv files, one for each degree. It also creates a degree_names_engineering.csv file on /data/YYYY-MM-DD/ that relates the degree name with the associated code.

Note: The YYYY-MM-DD states for the date where the information is retrieved.

The version that is currently on this repository calls the engineering() method, so you could download this code execute the script to download all the engineering alumni data into your computer:

python download_info_alumni_uc3m.py

That will ask you for your user and password so you can identify yourself in the UC3M servers. If you have any suspicion that the entering your credentials is not secure just take a look at the code, it's not rocket science.

Once I have all the data for the intended degrees (in this case the engineering ones), I want to combine them to create a unified file with all the data for all the degrees. The main reason for not creating this file in a first instance was to avoid loosing all the retrieved info if a connection fail occurs. By this implementation, the data is only saved if all the degree-related data has been retrieved. If the connection fails while retrieving the info, you can call again the function and it will start with the first degree that doesn't have any data (for a specific date).

I would also want to get rid of the names, because for the intended analysis is not really necessary and I wanted to make the data public without any personal information.

After that, there are some issues with the data that should be corrected (like the presence of non-engineering data and other inconsistencies).

For those tasks I have created the script transformdata.py. This script will create three additional csv file to the /data/YYYY-MM-DD/ directory:

• alumni.csv File with all the data of the intended degrees that are in the alumni/ folder.
• alumni_anonymous.csv File with all the data within the alumni.csv except for the names.
• alumni_anonymous_clean.csv Final version with all the corrected information from alumni_anonymous.csv.

As you can see, in this repository you can access to some of those csv files that I've mentioned on the /data/ folder.

Note that the directory you access doesn't have information about yourself, because its your account that's been used to connect to the system. In the csv files available above I have manually added information about myself, so the data should be consistent with all the alumni database (at that specific date).

Now that we have the data, we could proceed with any analysis we find interesting. My original purpose was to take a look of how much time take does it take to finish an engineering degree. With the previous described Python project we can get updated information about the different degrees that the UC3M has and how long did it take for each alumni within the database to graduate.

I have created an R script called generate_images_and_tables.R that makes a basic analysis of the engineering degrees. This script creates a bunch of histograms that can be viewed on the /doc/images/ folder, associating different type of studies in different images (degree, master, etc.). It also creates a summary table for each of those types (that can be consulted here.

My intention was also to publish this information in some way (maybe a blog or something similar) but by the time I decide how I will actuate, I have written a tiny file here describing the basic information of each category (in spanish).

Just to have a quick look, here is the corresponding analysis of the degree that I studied (Telecommunications Engineering - Ingeniería en Telecomunicación) among others of its category. These degrees are no longer available to study in Spain (now you have to study a BSc and a MSc to get the same qualification) and consisted on 5 courses (supposed to be 5 years):

• Ingeniería en Telecomunicación: Telecommunications Engineering
• Ingeniería Industrial: Industrial Engineering
• Ingeniería Informática: Computer Science

That means that more than 80% of the alumni for this type of degree finished their studies in more time than the stablished (5 years). In the case of my degree, more than 90% of the alumni needed more time.

It is also likely that most of the people (or probably all) that finished these degrees in less than 5 years is because they made only the last 2 courses (you could access the last two courses if you had a previous 3-year-course engineering degree).

Besides the general curiosity that I had, I also wanted to know a measure of the academic performance. Spanish universities usually don't give you any measurement of percentile once you finish (you basically don't know if you were one of the best), so I wanted to calculate some kind of performance comparing the years-to-finish with other fellow students of my degree (it would be great to know all the final marks to ponderate the results, but it's not the case).

For this case, I took the profiles for the graduated alumni for the last 5 years (graduating year from 2010 to 2014). I investigated those alumni that finished the degree in less than 5 years, and all of them had the previous 3-year-course engineering degree, so I didn't take them into account (I made a manual lookup on LinkedIn to check this out).

It turns out that only 5.5% have finished in 5 years, 9.5% have finished in 6 years and 12% have finished in 7 years. This means that, if you have finished within 6 years, like I did, you are in the top 15% with best performance.