[Verse 1] When labels are missing and you need to find The hidden patterns in data blind No teacher to guide you, no targets to chase Just structure waiting in feature space K-means will cluster your points in groups While DBSCAN finds density loops Hierarchical builds from bottom to top Gaussian mixtures where probabilities drop [Chorus] Unsupervised learning, finding what's there Clustering, reducing dimensions with care Silhouette scores and elbow method too Davies-Bouldin tells us what clusters can do No labels needed, just let data speak The patterns you find are the insights you seek [Verse 2] When dimensions are high and you can't visualize PCA finds the components that rise T-S-N-E maps the local neighborhood U-MAP preserves both local and global good Customer segments in e-commerce space Reduce then cluster to find their place Intrinsic metrics measure within Extrinsic compares to where you begin [Chorus] Unsupervised learning, finding what's there Clustering, reducing dimensions with care Silhouette scores and elbow method too Davies-Bouldin tells us what clusters can do No labels needed, just let data speak The patterns you find are the insights you seek [Bridge] Isolation Forest finds the outliers Autoencoders spot what's peculiar Local Outlier Factor checks the neighborhood Association rules find what's understood Market basket analysis shows what we buy When fraud detection needs to fly Apply these methods when structure's unclear And hidden relationships you want to hear [Verse 3] Evaluate wisely with the right technique Silhouette high is what you seek Davies-Bouldin lower is better for you Elbow method shows what K should do When exploratory work needs to start When preprocessing plays its part Choose unsupervised when patterns hide Let algorithms be your guide [Chorus] Unsupervised learning, finding what's there Clustering, reducing dimensions with care Silhouette scores and elbow method too Davies-Bouldin tells us what clusters can do No labels needed, just let data speak The patterns you find are the insights you seek [Outro] From customer segments to anomaly detection Unsupervised gives you data direction Apply the methods, evaluate right Hidden patterns come to light
← Unit 2.1 — Supervised Learning | Unit 2.3 — Reinforcement Learning →