Spatio-Temporal Clustering

import movekit as mkit
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib import animation

# Enter path to CSV file
path = "./datasets/fish-6.csv"

data = mkit.read_data(path)
data.head()

	time	animal_id	x	y
0	0	0	201.623002	186.148565
1	0	1	247.658670	177.650135
2	0	2	127.457230	20.572136
3	0	3	231.180722	172.861323
4	0	4	217.739517	189.203565

Movekit allows the use of many different clustering algorithms for spatio-temporal clustering: dbscan, hdbscan, agglomerative, kmeans, optics, spectral, affinitypropagation, birch.

#normalize the data first
data_norm = mkit.normalize(data)

#then cluster it
labels = mkit.clustering('dbscan', data_norm, eps1=0.05, eps2=10, min_samples=2)

#OR cluster with the splitting-and-merging method (data is partitioned into frames and merged afterwards).
#labels = mkit.clustering_with_splits('dbscan', data, frame_size=20, eps1=0.05, eps2=10, min_samples=3)

labels

array([0, 0, 1, ..., 3, 0, 0])

def plot(data, labels):
    colors = ['#a6cee3','#1f78b4','#b2df8a','#33a02c','#fb9a99','#e31a1c','#fdbf6f','#ff7f00','#cab2d6','#6a3d9a']

    for i in range(-1, len(set(labels))):
        if i == -1:
            col = [0, 0, 0, 1]
        else:
            col = colors[i % len(colors)]

        clust = data[np.where(labels==i)]
        plt.scatter(clust[:,0], clust[:,1], c=[col], s=1)
    plt.show()

    return None

data_np = data_norm.loc[:, ['time','x','y']].values
plot(data_np[:,1:], labels)