Skip to content

Repository with the code and report submitted to SUS7 Hackathon

Notifications You must be signed in to change notification settings

bergio13/sus2024-hackathon

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Stats Under the Stars 2024 - Hackathon

Our team composed of myself, Gianluca Procopio, Filippo Parlanti and Himel Ghosh achieved second place in the "Stats Under the Stars (SUS 7) Hackathon" an event endorsed by the Italian Statistical Society (SIS) and sponsored by Adecco and received the “Best Objective Prediction” award (€600). Stats Under the Stars (SUS) is a hackathon for Data Scientists, MSc and PhD students, where the teams have a full night to compete in the analysis of a real-world analytics problem. This year, SUS7 was organized by the University of Salento - Department of Human and Social Sciences in collaboration with Young SIS.

The challenge

The challenge consisted of solving a real-life problem. We were given 3 datasets: one for job vacancies, one for applicants, and one to join 5 different applicants to each job vacancy. Obviously, for each job only one candidate could be chosen while the same candidate could be chosen for multiple jobs.

Essentially, this was a classification problem that required a lot of preprocessing of the given datasets in order to obtain a final version to train our model on. We took advantages of similarities to try to figure out how similar a candidate was to a certain job position, such that for each couple Job-Candidate our model will produce a certain probability, and then for each job position we will set to 1 the candidate with the highest probability (i.e., the chosen one) and to 0 all the others.

On the test set we obtained an accuracy of 76%, scoring 2nd place in the leaderboard and winning the "Best objective prediction" prize as our solution striked the best balance between computational complexity and results.

Model Architecture and Results

Total Parameters: 9,729 (38.00 KB)
Trainable Parameters: 9,729 (38.00 KB)
Non-Trainable Parameters: 0 (0.00 Byte)

Layer (Type) Output Shape Parameters
Dense (128 units) (None, 128) 1,408
Dropout (30%) (None, 128) 0
Dense (64 units) (None, 64) 8,256
Dropout (30%) (None, 64) 0
Dense (1 unit) (None, 1) 65
Model Results

Prize

Links

About

Repository with the code and report submitted to SUS7 Hackathon

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published