Visualization
This visualization plots the Davies-Bouldin Index against the number of clusters (k) for clustered citation data from philosophy journals, showing that 7 clusters yield the lowest index and thus the best clustering. The scatterplot uses d3.svg.axis, d3.scale.linear, and d3.behavior.zoom for panning and zooming, with circles colored by cluster and tooltips via d3.tip. Data comes from a CSV file containing clustering results from the author's analysis.
AI-generated descriptionMethod
The 500 most cited or all accessible papers from the following journals where downloaded, and there citation records accessed:
analytical journals:
- Mind
- Analysis
- Philosophical Quarterly
- Ethics
continental journals:
- Continental Philosophy Review
- Journal of Speculative Philosophy
- Constellations
- Journal of the British Society for Phenomenology
The citation data was stripped to the cited authors. The resulting documents where tokenized - divided into single words. Then it was determined for every paper if it contained any of the found names.
| Paper Nr. | Heidegger | Ryle | ... |
|---|---|---|---|
| 1 | 1 | 0 | ... |
| 2 | 1 | 0 | ... |
| 3 | 0 | 1 | ... |
These resulting examples where clustered with a K-means-Algorithm. The Davies-Bouldin-Indexes for this Clustering are shown above. A low index indicates good clustering, which means that the clusters have high internal consistency and are very different to each other So we can say that the hypothesis of two clusters (eg. analytic and continental philosophy) seems not particulary merited (DB = 8.6), expecially when compared to 7 clusters, which yields a DB of 6.92.
forked from <a href='http://bl.ocks.org/MNoichl/'>MNoichl</a>'s block: <a href='http://bl.ocks.org/MNoichl/a92c39fd156938ba32bb08b3643c020d'>DB-Indexes for selected paper-clusters</a>