Long Short-Term Memory to Predict Unique Visitors of an Electronic Journal

Abstract

Unique visitors are visitors who use one IP in a certain period of time. The number of unique visitors every day is a benchmark for the success of an electronic journal. The increasing number of unique visitors every day shows that scientific periodicals are increasingly in demand by the wider community, which also affects the breadth of distribution, and speeds up the journal accreditation system. Therefore it is necessary to forecast the number of unique visitors in electronic journals in the future. Here, Long Short-Term Memory (LSTM) captures the pattern of data that has been obtained and then used to describe future data. The data used for testing is unique, ending data as of January 1, 2018, until December 31, 2018. After the data is obtained, the data will be normalized, then processed by the LSTM method to get the output. Then the output will be normalized to get the size of MSE, RMSE, and also the level of accuracy. The selection of the learning rate and the determination of the number of neurons in the LSTM process have an effect on the performance test performed. From the research conducted, the highest accuracy results obtained in the learning rate of 0.1 is 66.81%. While the lowest MSE and RMSE were obtained at a learning rate of 0.2 is 189.53 and 13.76. Thus, the results obtained are expected to be able to predict the number of unique visitors to electronic journals in the future to meet the needs of journal accreditation.