Two types of learning data were prepared for the SVM as follows.
(1) In 2011, Tsuji et al. (2011) asked 33 subjects to provide the name of one book that currently interested them.
Based on the books, they recommended 460 books using an association rule and they asked the subjects to evaluate
their degree of interest in the book that was recommended using the same criteria that are used in this study (to be
described later). In this way, they obtained 460 pairs of books. In this study, the pairs where the recommended
book was evaluated as “2: very interested” and “0: not interested” were considered as positive examples and
negative examples, respectively, and were used as the first set of learning data. The total number of pairs in the
learning set was 186 and the numbers of pairs of positive examples and negative examples were 59 and 127,
respectively.