The challenge is to find a better way of matching jobs and talents to opening/projects.
This was both to address the demand side: where an employee is given a project that she may not want or employer where they took time to consider multiple applicants only to find out that their best choice is taken by the time they return. (Worst, the applicants is also matched poorly as she feels that she needs to rush to accept the offer)
The inspiration for this comes from school matching system, although we developed a better algorithm to give an optimal match.
And the supply side: Where employees may not be aware of available jobs or the employer not being aware of potential employees. This also relates to finding the right employee given a new job that requires a combination of skills. Furthermore, with the algorithm for grouping into clusters, we are able to better plan employee skill progression and know where they are likely to venture into.
The inspiration for this comes largely from network analytics and association rules.
What it does
For the demand side, we assume that employee already knows what they want and are motivated to list their preference. What we do is to make it easier for them to list their true preferences and create an optimal match with the opening available.
For the supply side, we leverage on network analytics to give us key insights of employee skills (and miss-opportunity) where we could suggest to them to consider or for the employer to reach out to. We also managed to help cluster employee into natural groups which would lead to dynamic career paths and for HR to plan the skill to help these employees.
What more, we also help show for jobs that require a mix of very different skills, how to approach employee that are perhaps the best fit (despite not explicitly listed in their resume skillset)
How we built it
We build this using R with dummy data (assumed to be collected via employee surveys) We also use gephi to generate the network graphs and the associated statistics of a given network.
Challenges we ran into
We have a very rich discussion and generated lots of ideas. However, cutting down these ideas are sometimes a challenge. The dateline also places demands on the programming side of things, especially to ensure it runs error free. (although we are confident that we iron out all the bugs)
Accomplishments that we are proud of
We are proud to create a proof of concept, which is easily scalable to a large number of employee and job opening. (the current prototype already can handle any amount of employee/jobs)
What we learned
We learned a lot from the mentors, and what methods they have used. We are also surprised that HR was a challenge for accounting, and it made us rethink what we know of the sector.
What's next for Algorithms for job matching
We are now on the stage where we hope it could be tested in the real world. Perhaps getting feedback on how users find it as well as how to integrate into the HR work processes.
Technical issues to be hammered out are to generate dynamic filterings and better user experience flow.
We would be very happy to be able to work alongside industry experts to make this a reality.