def biclustering(db): #mydata = genfromtxt('/home/fan/intern/process_db/analysis/viewtime_matrix_524.csv',dtype=None,delimiter=',',names=True,skip_header=1) df = pd.read_csv( '/home/fan/intern/process_db/analysis/viewtime_matrix_501_0.1.csv') dma = 501 #print df.head() print df.shape dev_list = df.ix[:, 0].values prog_list = df.columns.values #print type(dev_list) #print type(prog_list) df.drop(df.columns[0], axis=1, inplace=True) #df[df==0] = 1 df = df.apply(fraction, axis=1) #print df.head() #print df.values #print type(df.values) #mydata = df.values #mydata=np.delete(mydata, 0, axis=0) #mydata=np.delete(mydata, 0, axis=1) #mydata[mydata==0] = 0.01 #print 'data format is:',mydata,type(mydata) # model=SpectralCoclustering(n_clusters=5, random_state=0) #n_clusters=(1000,20) # 4*3 = 12 clusters #model = SpectralBiclustering(random_state=None) model = SpectralCoclustering(n_clusters=10) model.fit(df) #fit_data=mydata[np.argsort(model.row_labels_)] #fit_data=fit_data[:,np.argsort(model.column_labels_)] #plt.matshow(fit_data[0:40],cmap=plt.cm.Blues) # plt.show() print model.get_params() for i in range(0, 5): print 'Size of one cluster:', model.get_shape(i) indices = model.get_indices(i) #print indices[1] print prog_list[indices[1]] print model.get_submatrix(i, df.values) dev_in_cluster = dev_list[indices[0]] #print type(dev_in_cluster) print 'number of devices within this cluster:', len(dev_in_cluster) get_income(db, dma, dev_in_cluster.tolist())
col_complement = np.nonzero(np.logical_not(cocluster.columns_[i]))[0] weight = X[rows[:, np.newaxis], cols].sum() cut = (X[row_complement[:, np.newaxis], cols].sum() + X[rows[:, np.newaxis], col_complement].sum()) return cut / weight bicluster_ncuts = list(bicluster_ncut(i) for i in xrange(len(newsgroups.target_names))) best_idx = np.argsort(bicluster_ncuts)[:5] print() print("Best biclusters:") print("----------------") for idx, cluster in enumerate(best_idx): n_rows, n_cols = cocluster.get_shape(cluster) cluster_docs, cluster_words = cocluster.get_indices(cluster) if not len(cluster_docs) or not len(cluster_words): continue # categories cluster_categories = list(document_names[i] for i in cluster_docs) counter = Counter(cluster_categories) cat_string = ", ".join("{:.0f}% {}".format(float(c) / n_rows * 100, name) for name, c in counter.most_common()[:3]) # words out_of_cluster_docs = cocluster.row_labels_ != cluster out_of_cluster_docs = np.where(out_of_cluster_docs)[0] word_col = X[:, cluster_words]
"""Items of a defaultdict(int) with the highest values. Like Counter.most_common in Python >=2.7. """ return sorted(iteritems(d), key=operator.itemgetter(1), reverse=True) bicluster_ncuts = list(bicluster_ncut(i) for i in range(len(newsgroups.target_names))) best_idx = np.argsort(bicluster_ncuts)[:5] print() print("Best biclusters:") print("----------------") for idx, cluster in enumerate(best_idx): n_rows, n_cols = cocluster.get_shape(cluster) cluster_docs, cluster_words = cocluster.get_indices(cluster) if not len(cluster_docs) or not len(cluster_words): continue # categories counter = defaultdict(int) for i in cluster_docs: counter[document_names[i]] += 1 cat_string = ", ".join("{:.0f}% {}".format(float(c) / n_rows * 100, name) for name, c in most_common(counter)[:3]) # words out_of_cluster_docs = cocluster.row_labels_ != cluster out_of_cluster_docs = np.where(out_of_cluster_docs)[0] word_col = X[:, cluster_words]
avg_data[row_sel, col_sel] = np.average(data[row_sel, col_sel]) avg_data = avg_data[np.argsort(model.row_labels_)] avg_data = avg_data[:, np.argsort(model.column_labels_)] plt.matshow(avg_data, cmap=plt.cm.Blues) plt.title("Average cluster intensity") plt.savefig('%s_averaged.png' % (identifier), bbox_inches='tight') if args.write: print "Writing clusters to database." # No need to clean up here, just overwrite by _id. for c in range(n_clusters): (nr, nc) = model.get_shape(c) (row_ind, col_ind) = model.get_indices(c) cluster_val = None if nr > 25 or nc > 50: print "Nulling cluster %d: shape (%d, %d)" % (c, nr, nc) else: cluster_val = c for ri in row_ind: data_list[ri]['cluster'] = cluster_val datastream.save(data_list[ri]) for ci in col_ind: events_list[ci]['cluster'] = cluster_val events.save(events_list[ci])