def dbscanClusterValidation(groundTruthMap, geneList, clusterList, noiseList,
file_name):
dbscanAlgoClusterMap = {}
size = len(geneList)
ExternalIndex_Computation.computeAlgoClusterMap(clusterList,
dbscanAlgoClusterMap,
noiseList)
groundTruthClusterMatrix = np.zeros((size, size), dtype=np.int)
algoClusterMatrix = np.zeros((size, size), dtype=np.int)
ExternalIndex_Computation.computeGroundTruthAndAlgoClusterMatrix(
groundTruthMap, geneList, groundTruthClusterMatrix,
dbscanAlgoClusterMap, algoClusterMatrix)
ExternalIndex_Computation.computeJaccardCoefficient(
groundTruthClusterMatrix, algoClusterMatrix, size)
lab = []
new_list = []
for i in range(0, len(clusterList)):
for j in range(0, len(clusterList[i])):
new_list.append(geneList[clusterList[i][j] - 1])
k = 1
for i in range(0, len(geneList)):
lab.append(dbscanAlgoClusterMap[k])
k += 1
no_of_clusters = len(np.unique(labels))
for i in range(len(labels)):
if labels[i] == -1:
no_of_clusters += 1
graph_label = "DBScan Result (" + file_name + ")"
pca(geneList, lab, graph_label, len(clusterList))
def createGeneList(file, groundTruthMap, geneList, file_name):
k=1
for line in iter(file):
record = line.strip().split("\t")
groundTruthMap[k] = int(record[1])
labels.append(int(record[1]))
k = k+1
floatRec =[]
for r in range(2,len(record)):
num = float(record[r])
floatRec.append(num)
geneList.append(floatRec)
no_of_clusters = int(np.unique(int(record[1])))
graph_label = "Hierarchical Original (" + file_name + ")"
pca(geneList, labels, graph_label, no_of_clusters)
def createInputMatrix(file, groundTruthMap, file_name):
geneList = []
k = 1
for line in iter(file):
record = line.strip().split("\t")
groundTruthMap[k] = int(record[1])
labels.append(int(record[1]))
k += 1
list = []
for i in range(2, len(record)):
list.append(float(record[i]))
geneList.append(list)
no_of_clusters = int(np.unique(int(record[1])))
#file.close()
graph_label = "DBScan Original (" + file_name + ")"
pca(geneList, labels, graph_label, no_of_clusters)
return geneList
def checkConvergence(final_centroid, list_final_clusters):
"""
Computes and checks if the new centroids are same as the old centroid. If yes, the algorithm converges; else
it again iterates over euclidean_distance function
.........................................................................................
Args:
{Input}: final_centroid, Computed new centroids of the data point
{Input}: list_final_clusters, List containing gene ids of every cluster
Returns:
{Output}: Computes the pca graph and passes values to compute external index results if the results
converges, else iterates from the euclidean distance
"""
#print("new cent.................",final_centroid)
#print("old...............", old_centroid)
answer = np.array_equal(final_centroid, old_centroid)
global count
if answer == False and no_of_iterations != count:
count +=1
old_centroid.clear()
for i in range(0, len(final_centroid)):
old_centroid.append(final_centroid[i])
final_centroid.clear()
euclidean_distance(brr, old_centroid)
else:
#print(count)
print("Cluster list: ", list_final_clusters)
graph_label = "K-means: Result("+ file_name+")"
j=1
labels_algo = []
for i in range(0, len(brr)):
for j in range(0, len(list_final_clusters)):
if i in list_final_clusters[j]:
labels_algo.append(j+1)
pca(brr, labels_algo,graph_label, no_of_clusters)
list_final_clusters1 = []
for i in range (0, len(list_final_clusters)):
templist = []
for j in range(0,len(list_final_clusters[i])):
templist.append(list_final_clusters[i][j]+1)
list_final_clusters1.append(templist)
kmeansClusterValidation(groundtruth_map,brr,list_final_clusters1)
def hcaClusterValidation(groundTruthMap, geneList, hcaClusters, file_name):
hcaAlgoClusterMap = {}
noiseList =[]
size = len(geneList)
ExternalIndex_Computation.computeAlgoClusterMap(hcaClusters, hcaAlgoClusterMap,noiseList)
groundTruthClusterMatrix = np.zeros((size,size), dtype=np.int)
algoClusterMatrix = np.zeros((size,size), dtype=np.int)
ExternalIndex_Computation.computeGroundTruthAndAlgoClusterMatrix(groundTruthMap, geneList, groundTruthClusterMatrix, hcaAlgoClusterMap,algoClusterMatrix)
ExternalIndex_Computation.computeJaccardCoefficient(groundTruthClusterMatrix,algoClusterMatrix,size)
no_of_clusters = len(hcaClusters)
graph_label = "Hierarchical Result (" + file_name + ")"
lab = []
new_list = []
for i in range(0, len(hcaClusters)):
for j in range(0, len(hcaClusters[i])):
new_list.append(geneList[hcaClusters[i][j] - 1])
for i in range(0, len(hcaClusters)):
for j in range(0, len(hcaClusters[i])):
lab.append(hcaAlgoClusterMap[hcaClusters[i][j]])
pca(new_list, lab, graph_label, no_of_clusters)
old_centroid = arr[initial_clusters][:,2:]
old_centroid = old_centroid.tolist()
# Get the all ground truth labels
labels = arr[:,1]
labels = labels.astype(np.int64)
global unique_lab
unique_lab = np.unique(labels) # All unique labels
unique_length = len(np.unique(labels))
for i in range(0,len(unique_lab)):
if unique_lab[i] == -1:
unique_length = len(unique_lab) -1
# Calculate pca of original data set
pca(brr, labels, graph_label, unique_length)
unique_length = no_of_clusters
#random = np.random.randint(0,arr.shape[0]-1, unique_length) # [378, 55, 51, 35, 237] [464,246,261,397,396,84,67,361,412,247,4], [505,364,145,456,60,41,472,366,460,449,4]
groundtruth_map = {}
k = 1
for i in range(0,len(arr)):
groundtruth_map[k] = int(arr[i][1])
k += 1
# Calculate euclidean distance
start_time = time.time()
euclidean_distance(brr, old_centroid)
print("--- %s seconds ---" % (time.time() - start_time))