def level2_candidate_gen(L, sortedItems, SDC, totalSequence, MIS, SequenceDataset):
Candidate2 = []
seq=[]
# Checks the combinations of items that had been discovered and checks the support requirement of the combinations.
for i in range(0, len(sortedItems)):
if sortedItems[i] in L.keys():
for j in range(0, len(sortedItems)):
if sortedItems[j] in L.keys():
# Checks level 2 candidate generation logic, including the SDC check using float data type.
if (float(L[sortedItems[i]])/totalSequence >= min(MIS[sortedItems[i]], MIS[sortedItems[j]]) and abs(float(L[sortedItems[i]])/totalSequence-float(L[sortedItems[j]])/totalSequence) <= SDC and float(L[sortedItems[j]])/totalSequence >= min(MIS[sortedItems[i]], MIS[sortedItems[j]]) ):
# Logic to form {a,b} item set
if sortedItems[i] < sortedItems[j]:
seq.append(sortedItems[i])
Candidate2.append(Transaction([[sortedItems[i], sortedItems[j]]]))
seq.append(sortedItems[j])
Candidate2.append(Transaction([[sortedItems[i]], [sortedItems[j]]]))
else:
continue
else:
continue
line9_16.line9_16(SequenceDataset, Candidate2, MIS)
return Candidate2
def MScandidate_gen_SPM(fkminus1, MIS, SequenceDataset):
Candidate = []
final_seq1_del= []
final_seq2_del = []
# Perform combinations of previously found frequent sets.
for i in range(0, len(fkminus1)):
for j in range(0, len(fkminus1)):
# Extract items of each sequence in consideration into each temporary variables.
sequence1 = copy.deepcopy(fkminus1[i].itemset)
sequence2 = copy.deepcopy(fkminus1[j].itemset)
seq1 = []
for item in itemInSequence(sequence1):
seq1.append(MIS[item])
seq2 = []
for item in itemInSequence(sequence2):
seq2.append(MIS[item])
# Checks of the MIS values of last item in sequence 2 is less than MIS value of every other element in S2
if seq2[-1] <= min(seq2[0:len(seq2)-1]):
# create temporary copy to store modified sequences
s1MinusFirst = copy.deepcopy(sequence1)
s2_dropseclast = copy.deepcopy(sequence2)
# Modify the first sequence
if len(sequence1[0]) > 1:
seq1_del = s1MinusFirst[0][0]
s1MinusFirst[0].remove(s1MinusFirst[0][0])
else:
seq1_del = sequence1[1:len(sequence1)]
s1MinusFirst = sequence1[1:len(sequence1)]
# Modify second sequence
if len(sequence2[len(sequence2)-1]) != 1:
seq2_del = s2_dropseclast[len(s2_dropseclast)-1][len(s2_dropseclast[len(s2_dropseclast)-1])-2]
s2_dropseclast[len(s2_dropseclast)-1].remove(s2_dropseclast[len(s2_dropseclast)-1][len(s2_dropseclast[len(s2_dropseclast)-1])-2])
else:
seq2_del = s2_dropseclast[len(s2_dropseclast)-2][len(s2_dropseclast[len(s2_dropseclast)-2])-1]
s2_dropseclast[len(s2_dropseclast)-2].remove(s2_dropseclast[len(s2_dropseclast)-2][len(s2_dropseclast[len(s2_dropseclast)-2])-1])
if seq1_del == seq2_del:
flag = True
# eliminates any empty sets that are discovered after deletion
s1MinusFirst=emptySetRemove(s1MinusFirst)
s2_dropseclast=emptySetRemove(s2_dropseclast)
seq1_temporary = []
seq2_temporary = []
# Checks if the two resulting sequences are equal
if equalityCheck(s1MinusFirst, s2_dropseclast) and (MIS[sequence2[-1][-1]] <= MIS[sequence1[0][0]]):
# Insert the values depending on the length of sequences
if len(sequence1[0]) == 1:
seq1_temporary.append(sequence1[0])
seq2_temporary.append(sequence2)
s2_tmp = copy.deepcopy(sequence2)
sequence2.insert(0, sequence1[0])
Candidate.append(Transaction(sequence2))
# if the length of sequence 2 is 2 then append the new values accordingly
if (len(s2_tmp) == 2 and elementLength(s2_tmp) == 2) and (float(sequence1[0][0]) < float(s2_tmp[0][0])):
s2_tmp[0].insert(0, sequence1[0][0])
Candidate.append(Transaction(s2_tmp))
# if sequence 2 is of length 1 and sequence 2 is of length 2 then append sequence 2
elif (len(sequence2) == 1 and elementLength(sequence2) == 2 and float(sequence1[0][0]) < float(sequence2[0][0])) or elementLength(sequence2) > 2:
sequence2[0].insert(0, sequence1[0][0])
Candidate.append(Transaction(sequence2))
continue
# Case 2: if the MIS values of 1st item in sequence 1 is less than every other item
if seq1[0] <= min(seq1[1:len(seq1)]):
# Create temporary place holders for modified sequence
s1_dropsecond = copy.deepcopy(sequence1)
s2MinusLast = copy.deepcopy(sequence2)
# Modify both the sequences
if len(sequence1[0]) > 1:
seq1_del = s1_dropsecond[0][1]
s1_dropsecond[0].remove(s1_dropsecond[0][1])
else:
seq1_del = s1_dropsecond[1][0]
s1_dropsecond[1].remove(s1_dropsecond[1][0])
if len(sequence2[len(sequence2)-1]) > 1:
seq2_del = s2MinusLast[len(s2MinusLast)-1][len(s2MinusLast[len(s2MinusLast)-1])-1]
s2MinusLast[len(s2MinusLast)-1].remove(s2MinusLast[len(s2MinusLast)-1][len(s2MinusLast[len(s2MinusLast)-1])-1])
else:
seq2_del = sequence2[0:len(sequence2)-1]
s2MinusLast = sequence2[0:len(sequence2)-1]
if seq1_del != seq2_del:
flag = False
else:
flag = True
# Eliminate empty itemsets if any after modifying the sequences
s1_dropsecond=emptySetRemove(s1_dropsecond)
s2MinusLast=emptySetRemove(s2MinusLast)
seq1_add = []
# Checks if the two resulting sequences are equal
if equalityCheck(s1_dropsecond, s2MinusLast) and (MIS[sequence2[-1][-1]] >= MIS[sequence1[0][0]]):
# Append candidates as per the length of existing sequence
if len(sequence2[-1]) == 1:
seq1_temp = copy.deepcopy(sequence1)
# append the values at the end of sequence 1
sequence1.append(sequence2[-1])
seq1_add.append(sequence2[-1])
Candidate.append(Transaction(sequence1))
# if both the lengths are same but last item of sequence 2 > last item of sequence 1
if (len(seq1_temp) == 2 and elementLength(seq1_temp) == 2) and (float(sequence2[-1][-1]) > float(seq1_temp[-1][-1])):
if flag == True:
seq1_del = seq1_temp
seq1_temp[-1].append(sequence2[-1][-1])
seq1_add.append(Transaction(sequence1))
Candidate.append(Transaction(seq1_temp))
# length of sequence 1 is 1 and length of sequence 2 is 2, and last item condition as above, or length of sequence 1 is greater than 1
elif (len(sequence1) == 1 and elementLength(sequence1) == 2 and float(sequence2[-1][-1]) > float(sequence1[-1][-1])) or elementLength(sequence1) > 2:
if flag == True:
seq1_del = seq1_temp
sequence1[-1].append(sequence2[-1][-1])
#seq1_add.append(Transaction(seq1_temp))
Candidate.append(Transaction(sequence1))
continue
# case 3: if none of the above, check for the join step
else:
# Create temporary place holders of both the sequences
s1MinusFirst = copy.deepcopy(sequence1)
s2MinusLast = copy.deepcopy(sequence2)
# Modify both the sequences
if len(sequence1[0]) > 1:
seq1_del = s1MinusFirst[0][0]
s1MinusFirst[0].remove(s1MinusFirst[0][0])
else:
seq1_del = s1MinusFirst[0][0]
s1MinusFirst = sequence1[1:len(sequence1)]
if len(sequence2[len(sequence2)-1]) != 1:
seq2_del = s2MinusLast[len(s2MinusLast)-1][len(s2MinusLast[len(s2MinusLast)-1])-1]
s2MinusLast[len(s2MinusLast)-1].remove(s2MinusLast[len(s2MinusLast)-1][len(s2MinusLast[len(s2MinusLast)-1])-1])
else:
seq2_del = sequence2[0:len(sequence2)-1]
s2MinusLast = sequence2[0:len(sequence2)-1]
# Check for empty itemsets in modified sequences
s1MinusFirst=emptySetRemove(s1MinusFirst)
s2MinusLast=emptySetRemove(s2MinusLast)
# Check of the modified sequences are equal
if equalityCheck(s1MinusFirst, s2MinusLast):
if len(sequence2[-1]) == 1:
sequence1.append(sequence2[-1])
seq1_add.append(Transaction(sequence1))
Candidate.append(Transaction(sequence1))
else:
sequence1[-1].append(sequence2[-1][-1])
seq1_add.append(Transaction(sequence1))
Candidate.append(Transaction(sequence1))
final_seq1_del.append(seq1_del)
final_seq2_del.append(seq2_del)
# Once the cases have been evaluated and temporary candidate list has been prepared, we proceed to the pruning step
seq_intermediate = {}
# Place holder for final candidates
finalCandidate = []
minSupp_seq = 100
#for each discovered candidate check the minimum support
for seq in range(0, len(Candidate)):
seq_intermediate[seq] = 1
minSupp_seq = Candidate[seq].minSup
subSequence = copy.deepcopy(Candidate[seq].itemset)
for i in range(0, len(subSequence)):
for j in range(0, len(subSequence[i])):
temp_seq = copy.deepcopy(subSequence)
temp_seq[i].remove(temp_seq[i][j])
temp_seq = emptySetRemove(temp_seq)
# check for the availability in sequence and minimum MIS match, if not, skip the candidate
if (not validateFrequent(temp_seq, fkminus1)) and (findMinMIS(temp_seq, MIS) == findMinMIS(subSequence, MIS)):
seq_intermediate[seq] = 0
#minSupp_seq = seq_intermediate[seq].minSup
break
# if match found, consider the candidate and move to next sub sequence
else:
continue
if (not validateFrequent(temp_seq, fkminus1)) and (findMinMIS(temp_seq, MIS) == findMinMIS(subSequence, MIS)):
break
# Once the availibility and MIN MIs has been evaluated, store the final candidates
for i in range(0, len(seq_intermediate)):
if seq_intermediate[i]:
finalCandidate.append(Candidate[i])
# Calculate the rest of the parameters like count, MIN MIS for each candidates
line9_16.line9_16(SequenceDataset, finalCandidate, MIS)
return finalCandidate, final_seq1_del, final_seq2_del, minSupp_seq
