KLASIFIKASI KELUHAN PELANGGAN BERBASIS TWITTER MENGGUNAKAN METODE SUPPORT VECTOR MACHINE (SVM) (STUDI KASUS POS INDONESIA)

WULANDARI, YANNY (2019) KLASIFIKASI KELUHAN PELANGGAN BERBASIS TWITTER MENGGUNAKAN METODE SUPPORT VECTOR MACHINE (SVM) (STUDI KASUS POS INDONESIA). Other thesis, Universitas Sebelas Maret.

[img] PDF - Published Version
Download (347Kb)

    Abstract

    Pos Indonesia is a state-owned company engaged in freight forwarding services utilizing Twitter as a customer service. Here the admin will answer and resolve customer complaints that mention or direct message to the @PosIndonesia twitter account manually. One of the weaknesses of submitting complaints through Twitter is that tweets are in the form of unstructured digital text, making it difficult to channel to existing problem areas to be addressed immediately. Not to mention the number of complaints and similar complaints from several customers. Based on these problems, this study aims to classify the data of tweets into several categories of complaints, namely delays, system errors, sending failures, guarantee of goods, service personnel, and speed of response. The method used in this research is feature extraction using TF-IDF and classification using Support Vector Machine (SVM) by comparing linear, polynomial, and RBF kernels and finding the best parameters for each test using 10-fold cross validation, then also the precision, recall, and f1-score will be sought using the confussion matrix. The results of the experiment show that the highest average accuracy value uses a linear kernel that is 81.26% followed by the RBF kernel 81.44% and the last polynomial 67.12%. While for precision, recall, and f1-score the highest value is using linear kernels which are 90%, 89%, and 89% respectively. So that it can be concluded that the complaints data on the tweets of the Pos Indonesia account can be classified properly using a linear kernel. Keyword- text classification, twitter, tf-idf, Support Vector Machine, linear, non linear

    Item Type: Thesis (Other)
    Subjects: Q Science > Q Science (General)
    Divisions: Fakultas Matematika dan Ilmu Pengetahuan Alam > Informatika
    Depositing User: Nisa' Khoirun
    Date Deposited: 09 Apr 2019 17:13
    Last Modified: 09 Apr 2019 17:13
    URI: https://eprints.uns.ac.id/id/eprint/43593

    Actions (login required)

    View Item