Classification Of Mobile Application User Ratings Based On Data From Google Play Store
Kiki Ahmad Baihaqi, Eko Sediyono, Christine Dewi, Indrastanti R. Widiasari, Ahmad Fauzi

Universitas Kristen Satya Wacana
Universitas Buana Perjuangan Karawang

Abstract

Apart from using application acceptance methods, mobile application assessments also include assessments and comments which are part of the assessment of mobile-based service providers, especially those with Android operating systems, from this data, the data can be mined and then negative or positive results can be formulated. The good and bad aspects of the application are immediately assessed in real-time when the application is used. Several algorithms can conclude data scraped from the website. So, in this research, we prove with data like this which algorithms are good and optimal and which are good but not optimal. The results of this research are that many of the comments were positive or had parameters above 3, while only 15.3% had ratings below three with the data taken, namely data from the 1000 most recent comments. The three algorithms, namely logistic regression, support vector machine, and K-Nearest Negborn with each result sequentially, namely from an accuracy of 84 89 88 in percentage units. Then 87 86 91 precision percentage, and finally the F1 score is 65 80 76. The result of all this is that in this case SVM is more suitable and has a better score.

Keywords: Data Mining, Classification, KNN, SVM, Logistic Regression

Topic: Life Sciences