Exemple #1
0
import utilities


REUTERS_DIRECTORY = "/home/0/srini/WWW/674/public/reuters"

# Change num_files to a sample size of between 1-21 reuters files
vector_dataset = utilities.preprocessData(directoy=REUTERS_DIRECTORY, num_files=5)

just_body = vector_dataset["words_vectors"]
topic_and_body = vector_dataset["words_and_topics_vectors"]
ground_truth_labels = vector_dataset["topics_classes"]

# metric can be any in scipy.spatial.distance module
# i.e.  'minkowski', 'euclidean', 'dice', 'jaccard', 'cosine', etc.
# num_means should be < 50 for a sample of under 8 files or an error may be thrown
k_results = utilities.kmeans_cluster(topic_and_body, ground_truth_labels, num_means=40, metric="euclidean")
# k_results = utilities.kmeans_cluster(just_body, ground_truth_labels, num_means=40, metric='euclidean')

kclusters = k_results["clusters"]  # to see clustered labels in an interactive console


# linkage can be 'complete','average' 'ward' (only for euclidean)
# metric can be 'cosine', 'manhatten', 'euclidean'
h_results = utilities.hierarchical_cluster(
    topic_and_body, ground_truth_labels, number_of_leafs=40, linkage="complete", metric="euclidean"
)
# h_results = utilities.hierarchical_cluster(just_body, ground_truth_labels, number_of_leafs=leafs, linkage='complete', metric='euclidean')

hclusters = h_results["clusters"]  # to see clustered labels in an interactive console
Exemple #2
0
#!/usr/bin/env python
import utilities


REUTERS_DIRECTORY = "/home/0/srini/WWW/674/public/reuters"
#REUTERS_DIRECTORY = "../reuters"

# Change num_files to a sample size of between 1-21 reuters files
vector_dataset = utilities.preprocessData(reuters_directory=REUTERS_DIRECTORY, num_files=8)

just_body = vector_dataset["words_vectors"]
topic_and_body = vector_dataset["words_and_topics_vectors"]
ground_truth_labels = vector_dataset["topics_classes"]
int_labels = vector_dataset["topics_ints"]


# utilities.knn_classify(topic_and_body, int_labels, 0.5)
# utilities.naive_bayes_classify(topic_and_body, int_labels, 0.5)
# utilities.dtree_classify(topic_and_body, int_labels, 0.5)


utilities.knn_classify(topic_and_body, int_labels, 0.66)
utilities.naive_bayes_classify(topic_and_body, int_labels, 0.66)
utilities.dtree_classify(topic_and_body, int_labels, 0.66)


# utilities.knn_classify(topic_and_body, int_labels, 0.8)
# utilities.naive_bayes_classify(topic_and_body, int_labels, 0.8)
# utilities.dtree_classify(topic_and_body, int_labels, 0.8)