Implementasi Vector Space Model dalam Pembangkitan Frequently Asked Questions Otomatis dan Solusi yang Relevan untuk Keluhan Pelanggan di UPT PUSKOM UNS

摘要

UPT PUSKOM UNS as an service unit needs customer’s complaint handling. Customer’s complaints will be given solutions based on the past complaints which has similarity with the new complaints. Therefore, a method to calculate similarity between new complaint dan the past complaints is needed. The result of the calculation can be used for generating automatic Frequently Asked Questions (FAQ) and relevant solutions. There are some methods can be used for calculating document similarity, such as VSM. VSM is a method that has efficient procedure, easily represented dan can be implemented in document-matching. Therefore, in this research VSM in generating automatic FAQ and relevant solutions for customer’s complaint in UPT PUSKOM UNS will be used. Weighting term usedTerm Frequency-Inverse Document Frequency (TF-IDF) technique. Compared combinations are TF-IDF it self, logarithmic modified TF and logarithmic modified IDF. Similarity measure used cosine similarity. The results of this research are VSM algorithm with TF-IDF weighting can be used to generate automatic FAQ and the relevant solutions. Based on the accuracy calculation of each experiment can be concluded on a threshold 0.5, the combination of TF-IDF notation which has an average rating of highest accuracy and precision is TF-IDF, that is respectively 62.09% and 55.15%. Whereas in the threshold 0.65 that has average rating of the highest accuracy and precision is the first modification, which is respectively 83.18% and 68.35%. Besides that, the experiment using 171 data TF-IDF and threshold 0.65 can generate 27 FAQ, that is percentage70.37% is relevant.