diff options
Diffstat (limited to 'README.md')
| -rw-r--r-- | README.md | 41 |
1 files changed, 32 insertions, 9 deletions
@@ -3,9 +3,10 @@ K-Means Clustering [K-Means][kmeans] clustering implementation. Features: -* Initialization methods: random, [forgy][forgy], and [k-means++][kmeanspp]. -* Uses [Silhouette method][silhouette] to determine the optimum cluster - count. +* Initialization methods: random, [forgy][forgy], and + [k-means++][kmeanspp]. +* Uses the [Silhouette method][silhouette] to determine the optimum + cluster count. * No external dependencies. Examples @@ -18,15 +19,19 @@ Generate 1000 rows of test data, clustered around 3 points: # classify data, plot results as png ./km-test kmeans example.dat example.png -Run all tests and save best clustering in current directory along with +Run all tests and save best results in current directory along with PNGs of results: for i in tests/*dat; do - # build paths - png_path=kmeans-$(basename ${i/dat/png}) + # path to output data file + # (ex: src: "tests/c3-1e2-0.dat", dst: "kmeans-c3-1e3-0.dat") dst_path=kmeans-$(basename $i) - # run test + # path to output png + # (ex: src: "tests/c3-1e2-0.dat", dst: "kmeans-c3-1e3-0.png") + png_path=kmeans-$(basename ${i/dat/png}) + + # run test (use kmeans for initialization) echo $i ./km-test kmeans $i $png_path > $dst_path done @@ -42,6 +47,9 @@ Supported initialization methods: to pick the initial cluster centroids. This is the recommended initialization method. +It's probably best to just stick with [kmeans][kmeanspp] unless you know +what you're doing. + Data File Format ---------------- Reads and writes newline-delimited plain text files in the following @@ -66,8 +74,23 @@ Example data file: 9.8 6.5 4.3 3.2 5.6 8.7 -See the files in `tests/` for additional examples. You can also use -the top-level `gen-data.rb` script to generate additional test data. +See the files in `tests/` for additional example data files. You can +also use the top-level `gen-data.rb` script to generate additional test +data. + +Test Data +--------- +The test data files in the `tests/` directory use the following naming +convention: + + c<num_clusters>-<num_rows>-<N>.dat + +Where: + +* `num_clusters`: Number of clusters. +* `num_rows`: Number of rows, in exponent notation (`1e3` = 1000, `1e4` + = 10000, etc). +* `N`: Distinguishing suffix (usually 0). [kmeans]: https://en.wikipedia.org/wiki/K-means_clustering "K-Means clustering" [kmeanspp]: https://en.wikipedia.org/wiki/K-means%2B%2B "k-means++ initialization method" |