# -*- coding: utf-8 -*-

import numpy as np
import matplotlib.pyplot as plt
from sklearn.metrics import adjusted_rand_score
from sklearn.model_selection import GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans
from sklearn.cluster import SpectralClustering, AgglomerativeClustering

def predictWithK(testFeatures, numVessels, trainFeatures=None,
            trainLabels=None):
# Unsupervised prediction, so training data is unused

scaler = StandardScaler()
testFeatures = scaler.fit_transform(testFeatures)
#km = KMeans(n_clusters=numVessels, random_state=100)
#clf = AgglomerativeClustering(n_clusters=numVessels, linkage='average')
clf = SpectralClustering(n_clusters=numVessels,
                        random_state=42,
                        n_neighbors=20,
                        affinity='nearest_neighbors',
                        eigen_solver='arpack',
                        assign_labels='kmeans')
predVessels = clf.fit_predict(testFeatures)

return predVessels

def predictWithoutK(testFeatures, trainFeatures=None, trainLabels=None):
# Unsupervised prediction, so training data is unused

# Arbitrarily assume 20 vessels
return predictWithK(testFeatures, 5, trainFeatures, trainLabels

1. The decision boundary in any binary classification problem can b...