In the library, it has been a hard time for librarians to decide whether to purchase the replacement for a book that has been reported missing since some books may be found not long after the replacements have been bought. On the other hand, it would be irritating for us as students if books we’re looking for have been report missing for a long time without replacements.
As a group of students who wish to make school better, our goal is to help the NTHU library identify which missing book is likely to be gone forever and which is very likely to be found, so they can decide whether to initiate finding a replacement. If we are successful, we would not only eliminate the waste on redundant spending, we would also contribute to enhancing the efficiency of NTHU library administration.
Data mining goal
We will attempt to identify which (kinds of) books once missing are most likely to remain missing. We will attempt to produce: a binary classification of books - forever missing or will be found; and a ranking by likelihood to remain missing. This is both a supervised and predictive task, as the records made available to us contain data showing whether the current status of the book (lost or found). Additionally, the solution we hope to produce can be applied both retrospectively and prospectively - the library can use our solution to classify books in its larger record depending on our current status now and in the future. The main outcome variables is “current item status”