Syarip Saputra, Muhamad Fakhri Hidayat, Wahyu Gunawan, Indra Riyana Rahadjeng

KLASIFIKASI DAMPAK KECANDUAN MEDIA SOSIAL MAHASISWA DENGAN SVM DAN K-MEANS

Introduction

Klasifikasi dampak kecanduan media sosial mahasiswa dengan svm dan k-means. Klasifikasi dampak kecanduan media sosial mahasiswa menggunakan model hybrid SVM & K-Means untuk sistem peringatan dini. Akurasi 93.62% efektif untuk deteksi preventif di perguruan tinggi.

Abstract

Penggunaan media sosial yang meningkat di kalangan mahasiswa, disertai pola tidur tidak teratur, menjadi permasalahan serius karena berpotensi menurunkan prestasi akademik dan sering terlambat terdeteksi oleh institusi pendidikan. Penelitian ini bertujuan mengembangkan model prediksi hybrid dua tahap untuk mengklasifikasikan dampak kecanduan media sosial terhadap mahasiswa sebagai sistem peringatan dini. Metode yang digunakan mengintegrasikan algoritma K-Means untuk pelabelan tingkat risiko secara otomatis dan Support Vector Machine (SVM) sebagai tahap akhir klasifikasi. Penelitian menggunakan 705 data responden mahasiswa yang diproses melalui tahap preprocessing. Jumlah cluster optimal ditentukan menggunakan metode Elbow dan Silhouette Score, sedangkan kinerja model dievaluasi menggunakan confusion matrix. Hasil penelitian menunjukkan bahwa K-Means membentuk tiga klaster risiko (rendah, sedang, dan tinggi) dengan nilai Silhouette Score sebesar 0,4188. Model SVM menghasilkan akurasi 93,62%, presisi 94,62%, dan recall 90%, sehingga efektif mendukung pengambilan keputusan preventif di perguruan tinggi.

Review

The study, "KLASIFIKASI DAMPAK KECANDUAN MEDIA SOSIAL MAHASISWA DENGAN SVM DAN K-MEANS," addresses a timely and critical issue concerning the increasing prevalence of social media addiction among university students and its potential impact on academic performance and irregular sleep patterns. The authors propose an innovative two-stage hybrid predictive model, integrating K-Means for automated risk level labeling and Support Vector Machine (SVM) for final classification, to serve as an early warning system. The clear objective, coupled with a robust dataset of 705 student responses, positions this research as a significant contribution to understanding and mitigating the negative consequences of social media overuse in academic settings. The methodology is well-articulated, detailing the application of K-Means to identify three distinct risk clusters (low, medium, high), with optimal cluster determination informed by Elbow and Silhouette Score methods. The reported Silhouette Score of 0.4188 indicates a reasonable, though moderate, separation of these clusters. The subsequent SVM classification stage demonstrates impressive performance metrics, achieving an accuracy of 93.62%, precision of 94.62%, and recall of 90%. These high scores, evaluated using a confusion matrix, suggest the model's strong capability in accurately classifying the impact of social media addiction, thereby effectively supporting proactive decision-making in higher education institutions. While the study presents compelling results, further clarity on the specific features or variables used to define and quantify "dampak kecanduan media sosial" beyond "pola tidur tidak teratur" would enhance the replicability and interpretability of the model. Additionally, exploring the generalizability of these findings across different university contexts or cultural backgrounds could be a valuable avenue for future research. Nevertheless, this paper offers a practical and high-performing analytical tool that holds substantial promise for identifying at-risk students and enabling timely interventions, thus making a commendable contribution to student well-being and academic support strategies.

Full Text

Comments

You need to be logged in to post a comment.