predict method

Lesson 11 of 19 in Coddy's Introduction to Machine Learning course.

Challenge

Medium

The predict method calculates the distance of every single new point to all other centroids. The point is assigned to the cluster of the closest centroid.

For each new point, assign the index of the closest centroid. return the list of predicted assigned clusters.

Try it yourself

def euclidian_distance(point_a, point_b):
    return (sum([(point_a[i] - point_b[i])**2 for i in range(len(point_a))]))**0.5

class KMeans:
    def __init__(self, k):
        self.k = k

    def find_cluster_centroid(self, cluster_points):
    # Write your code here
        centroid = []
        for point in cluster_points:
            for dim, value in enumerate(point):
                if len(centroid) - 1 < dim:
                    centroid.append(0)
                centroid[dim] += value
        for i in range(len(centroid)):
            centroid[i] /= len(cluster_points)
        return centroid

    def fit(self, X_train):
        # Write you code here
        self.centroids = [X_train[i] for i in range(2, self.k*2 + 1, 2)]
        while True:
            # Assing each point to centroid
            points_clusters = []
            for point in X_train:
                distances = []
                for centroid in self.centroids:
                    dist_to_point = euclidian_distance(centroid, point)
                    distances.append(dist_to_point)
                closest_centroid_index = distances.index(min(distances))
                points_clusters.append((point, closest_centroid_index))

            # Update centroids
            clusters = {}
            for point, cluster in points_clusters:
                if cluster not in clusters:
                    clusters[cluster] = []
                clusters[cluster].append(point)
            new_centroids = [None]*self.k
            for cluster, points in clusters.items():
                new_centroids[cluster] = self.find_cluster_centroid(points)

            # Check if converged
            total_dist = 0
            for i in range(len(new_centroids)):
                total_dist += euclidian_distance(new_centroids[i], self.centroids[i])
            if total_dist < 0.01:
                break
            self.centroids = new_centroids


    def predict(self, X_test):
        pass