SUMMARY
The implementation of network security infrastructure has been carried out, including the Intrusion Detection System (IDS). However, in its implementation there are still many who have not combined with Data Technology (Data Science) to get a more comprehensive analysis. This study aims to analyze the types and characteristics of network threats using data science. As a computational method, the results of 3 algorithms in the unsupervised learning category will be implemented and compared, namely K-Means, Meanshift, and Density-Based Spatial Clustering of Applications with Noise (DBSCAN). From the experimental results as measured by the Silhouette Index (SI ) the best cluster of each implemented algorithm is DBSCAN which has the best SI value of 0.3424 with an Eps value of 0.2 and a MinPts value of 3. Meanwhile, from the results of clustering using K-Means, The best SI value was obtained by experiment k=4 with a value of 0.4531. The results of clustering using MeanShift, the best SI value was obtained by experiment bandwidth = 1 with a value of 0.5305.