Internet is a Media of Information and Communication Technology that is growing rapidly and in great demand by the public. This is evident from the number of users that increase each year. Internet is used by adults and children. The result of the national socio-economic survey 2016 showed that 6.91% of children aged 5-12 years old accessed the Internet. The ease of accessing the Internet becomes a problem when children use the Internet excessively. Increasing the number of children as Internet users need to be monitored to overcome the adverse impact of Internet use. One way is to know the characteristics of children as Internet users. Classification is an operation that places objects on a particular class based on its characteristics. In this study, we used a classification tree to form the classification of children as Internet users category of 5-12 years. We used data from the National Social Economic Survey surveyed by the Central Agency on Statistics in 2016 in Indonesia. The imbalance of data caused the insensitivity of the resulting classification to minority data. Handling imbalance data was applied using oversampling and under sampling.The objectives of this study are to determine the characteristics of children as Internet users, to see the effect of oversampling and under sampling, and to see the results of classification accuracy. The result was oversampling and under sampling increase sensitivity about 45%. Based on classification tree, it was known that children of Internet users were characterized by children who live in households with Internet expenditure of at least Rp.100.000 per month with many household members accessing the Internet.
Published In:IJCSN Journal Volume 7, Issue 1
Date of Publication : February 2018
Pages : 22-26
Figures :01
Tables : 02
Irene Imelda Juliaty Silaban : Statistics Department, Bogor Agricultural University Bogor, 16680, Indonesia.
Bagus-Sartono : Statistics Department, Bogor Agricultural University Bogor, 16680, Indonesia.
Indahwati : Statistics Department, Bogor Agricultural University Bogor, 16680, Indonesia.
Classification Tree, Internet, Oversampling, Under sampling
Dividing data into training and testing data was the first step before building a classification tree, with the proportion of training data as much as 75% and 25% of testing data amount to 117,576 and 39,192, respectively. The accuracy value using training data obtained sensitivity value of 33.22%, specificity of 98.90% and accuracy of 94.26%. The small sensitivity value means that the model of classification tree produced is not good enough to be used to classify data of children as Internet users.
[1] Ameliola S, Nugraha HD. 2013. Perkembangan media informasi dan tekonologi terhadap anak dalam era globalisasi. Diakses dari http://icssis.files.wordpress.com/2013/09/2013-0229 pada tanggal 10 September 2017. [2] [APJII] Asosiasi Penyelenggara Jasa Internet Indosia.2015. Profil Pengguna Internet Indonesia 2014. Jakarta (ID): Pusat Kajian Komunikasi Universitas Indonesia. [3] [APJII] Asosiasi Penyelenggara Jasa Internet Indosia.2015. Profil Pengguna Internet Indonesia 2014. Jakarta (ID): Pusat Kajian Komunikasi Universitas Indonesia. [4] Breiman L, Friedman JH, Olshen RA, Stone CJ. 1993. Classification and Regression Trees. New York, NY: Chapman and Hall. [5] Izenman AJ. 2008. Modern Multivariate Statistical Techniques. Department of Statistics Temple University Speakman Hall, Philadelphia, USA [6] Rahayu S, Adji TB, Setiawan NA. 2017. Analisis perbandingan metode over-sampling adaptive synthetic-nominal (ADASYN-N) dan adaptive synthetic-KNN (ADSYN-KNN) untuk data dengan fitur nominal-multi categories. 2017 Juli 27; Yogyakarta, Indonesia. CITEE. 296.
