Problem. Understand complex, high-dimensional datasets by uncovering structure and similarity through unsupervised visualization techniques.

Data.

• Grain samples — geometric seed features.
• Stock movements — daily price changes (2010–2015).
• Voting countries — similarity in international voting patterns.

Approach.

• Applied hierarchical clustering (single/complete linkage) to visualize merging patterns via dendrograms.
• Used t-Distributed Stochastic Neighbor Embedding (t-SNE) to map high-dimensional relationships into 2D space.
• Compared visual clusters with known categories (grain varieties, stock sectors, or country groups).

Results (qualitative).

• Hierarchical clustering revealed a meaningful grouping structure.
• t-SNE highlighted well-separated clusters, visually confirming similarity patterns.
• Showcased how scaling and linkage choices affect the shape and density of clusters.

What I Learned.

• Hierarchical vs. t-SNE perspectives on high-dimensional data.
• Reading dendrogram heights and linkage thresholds.
• How t-SNE’s stochastic nature impacts visualization stability.

git clone https://github.com/Joe-Naz01/t-sne_cluster.git
cd t-sne_cluster
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt
jupyter notebook