<![CDATA[Prediction of customer needs is needed to help ICON+ management provide maximum and quality ICONNET services to customers, and is useful for company management in planning related products offered, as well as providing input to management regarding products that are in great demand by customers. This study implements and compares the K-Nearest Neighbor (K-NN) and Support Vector Machine (SVM) algorithms in predicting the service Iconnet products that are most in demand by customers, so as to facilitate ICON+ management in planning customer service provision, and preparing market strategies in the future. future. A total of 3,206 datasets (consisting of 2,565 training data and 641 testing data) ICONNET service enthusiasts for 1 year that have been cleaned, were tested on both algorithms, based on parameters Bandwidth, Request_Date, Status, customer address, and service fee. Algorithm performance accuracy was tested using Cross Validation, Confusion Matrix and ROC curve methods. The results of the accuracy test show that the performance of the K-NN algorithm is more accurate than the SVM algorithm in various test categories.Keywords: Data Mining; Customer interest; Performance Accuracy; Cross Validation; Confusion Matrix Abstrak. Prediksi kebutuhan pelanggan diperlukan untuk membantu manajemen ICON+ menyediakan layanan ICONNET secara maksimal dan berkualitas kepada pelanggan, serta berguna bagi jajaran manajemen perusahaan dalam melakukan perencanaan terkait produk yang ditawarkan, serta memberi masukan pada pihak manajemen mengenai produk yang banyak diminati oleh pelanggan. Penelitian ini mengimplementasikan dan membandingkan algoritme K-Nearest Neighbor (K-NN) dan Support Vector Machine (SVM) dalam memprediksi layanan produk Iconnet yang paling diminati oleh pelanggan, sehingga dapat mempermudah manajemen ICON+ dalam perencanaan penyediaan layananan pelanggan, dan penyusunan strategi pasar di masa mendatang. Sebanyak 3.206 dataset (terdiri atas 2.565 data training dan 641 data testing) peminat layanan ICONNET selama 1 tahun yang telah dibersihkan, diuji pada kedua algoritme tersebut, berdasarkan parameter Bandwith, Request_Date, Status, alamat pelanggan, dan biaya layanan. Akurasi kinerja algoritme diuji menggunakan metode Cross Validation, Confusion Matrix serta ROC curve. Hasil uji akurasi menunjukkan kinerja algoritme K-NN lebih akurat dari algoritme SVM pada berbagai kategori pengujian.Kata kunci: Data Mining; Minat pelanggan; Akurasi Kinerja; Cross Validation; Confusion Matrix ]]>
