KLASIFIKASI PEMINJAMAN BUKU MENGGUNAKAN NEURAL NETWORK BACKPROPAGATION

Abstract

Book lending is a form of service provided by the library. Borrowing is very closely related to inventory. In library A in determining the inventory of books, library staff has difficulty in determining what types of books are needed by students. Where the determination of the number and type of books has not used a definite calculation system, only based on the estimated number of students and courses of each study program. Therefore a classification of borrowing types of books is needed based on student book lending transactions, the data used in this study were 269,116 from 2014 to 2019 months 6. The types of books processed were 184 types of books from 1700 types of books, the first step carried out perform forecasting techniques to predict the target inventory of each type of book in the following year, the second stage of the book lending transaction data is processed to determine the classification of book types using neural network backpropagation. Obtained the results of the error rate or MSE of 0.021, using hidden layer 9 and the activation function tansiq with epoch 2000, with the recommended number of book types recommended for restocking as many as 86 types of books with the number of predictions in each type of book. The third stage is to test the data validation to determine the level of classification and prediction errors, the last step is a regression test that shows a significant correlation of 0.006 with the data variables being tested namely prediction, classification and target data. The results of this study can provide book type recommendation data. along with the number of predictions in each type of book needed in the coming year using the neural network backpropagation method with an accuracy rate of 95.5%.