Library Optimization with Neural Networks

Optimizing book rental time with Machine Learning, to improve academic life and decrease library expenditure.

Comment

Inspiration

Every time I go to the library and want to borrow a book, I find very frustrating that despite having so many books lying on the shelf the give it to me only for 2 hours. And also certain books are very limited in quantity, but never available, because they are allowed to be rented out for many months. That is very wasteful in terms of resources, from both sides(students and library).

What it does

Our program takes a list of important attributes of a book, and calculates the optimal rental time. By optimal rental time we mean the time which minimizes the time books are lying in the library and maximizes the time students have the books.

Challenges we ran into

We had a hard time finding the data set for our neural network(not on the internet and no access to local library DB). That is why we came to the decision to generate our own data to show important attribute correlation.

How we built it

This problem is typical machine learning optimization task. The hardest part was generating good, relevant and realistic data. We accomplished it by a complex combination of many Gaussian functions on a different scale and with different randomized parameters. On top of that, we added noise, all this without breaking preset correlations between attributes.

The attributes we considered important for this task: book type(Science book, Fiction book, etc.), author, week in the year, day in the week, hour in the day, book id. This model is not complicated because of challenges we ran into during data generation, with additional time we considered attributes like: Rating of the book, Price of the book, Number of books in the library, Number of pages in the book, Publish date, etc.

Accomplishments that we're proud of

That we were able to find a solution model for important problem of optimization with usage of neural networks.

What we learned

Learned to apply theoretical knowledge in real life problems.

What's next for Library Optimization with Neural Networks

Getting access to real data and checking our assumptions about book parameters correlations. Add a lot more parameters to our model to further improve the accuracy. Find funding for further development or creation of public API which any library can use freely.

Built With

Share this project:

Updates