SimKit

Plug-in for Neo4j containing user-defined procedures to perform Spectral Clustering through a single procedure or as individual procedures (Similarity Graph, Laplacian Eigendecomposed Graph, KMeans). Additional procedures available for evaluation with Average Silhouette Score and Adjusted Rand Index, visualization of matrices into String and saving them to CSV.

About

Procedure Stack

Init SimKit

return simkit.initSimKit('bolt://localhost:7687', 'neo4j', '123412345')

Create Graph From Nodes

return simkit.nodePropertyToGraph({
    label: "Iris",
    distance_measure: "euclidean",
    graph_type: "full",
    parameter: 7,
    remove_column: "index,target"
})

Create EigenGraph

return simkit.nodePropertyEigen({
    label: "affinity_full_7_Iris",
    laplacian_type: "sym",
    number_of_eigenvectors: 3
})

Kmean Clustering

return simkit.kMeans({
    label: "eigen_sym_3_affinity_full_7_Iris",
    number_of_centroids: 3,
    number_of_iterations: 100,
    distance_measure: "euclidean",
    original_set: "Iris",
    overlook: "target,sepal_length,sepal_width,petal_length,petal_width",
    overlook_original: "target",
    silhouette: false
})

Calculate Adjusted rand index

return simkit.adjustedRandIndex({
    label: "Iris",
    true_labels: "target"
})

Spectral Clustering

return simkit.spectralClustering({
    label: "Iris",
    is_feature_based: true,
    distance_measure: "euclidean",
    graph_type: "full",
    parameter: 7,
    remove_columns: "index,target",
    laplacian_type: "sym",
    number_of_eigenvectors: 3,
    number_of_iterations: 100,
    distance_measure_kmean: "euclidean",
    target_column: "target",
    silhouette: false,
    seed: 42
})

Visualize Matrix and save as CSV

return simkit.getMatrix("affinity_full_7_Iris","adjacency","/path/to/folder")

Changelog

• Renamed main. to simkit.
• added initSimkit as initialisation function to remove hardcoded login creds
• Changed test Cases to enable compilation on all devices