DMML ASSIGNMENT
                                              APRIORI ALGORITHM

ABOUT: Created a program in python3.7 to compute frequent itemsets using apriori algorithm (both fixed minimum support and multiple minimum support approach)from the data sets docword.enron.txt ,dockword.nips.txt ,dockword.kos.txt availavle at

http://archive.ics.uci.edu/ml/datasets/Bag+of+Words .

PARAMETERS:

For fixed minimum support approach:

parameters used :
      kos:
          F=0.2,0.25,0.3,0.35,0.4,0.45,0.5,0.6,0.7 and 0.1
      nips:
          F=0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,0.95,0.98
      enron:
            ((N.B:The program didn’t run on pc. Reason:Pivot of Data set too big(8.3 G.B) to load for our program.So, ran in Google colab))
	  F=0.8,0.7,0.6,0.5,0.4,0.3,0.2,0.15,0.1

For multiple minimum support approach:

parameters used:
     kos:
         F=max of ({(Total no of occurances/total no document ) for each 
		     wordId},0.05)
         phi=0.001
	 k=2,3,4,5          
     nips:
         F=max of ({(Total no of occurances/total no document ) for each 
		     wordId},0.01)
         phi=0.006,0.6,0.06,0.02,0.0001
	 k=10

PROGRAM DESCRIPTION: LIBRARIES USED: numpy,pandas,time

FIXED MINIMUM SUPPORT APPROACH:

             candidate-gen function(candidate_gen(f(k-1))):
                         Takes input fk-1 returns the candidate set(ck) for computing fk
             counting support of a set(support_count_of_sets(input_set)):
                          counts the frequency of a candidate set                            
             Apriori function(apriori(data_clean,min_sup,k):
                  Takes input data, fixed minimum support and k(max length of frequent item set)

MULTIPLE MINIMUM SUPPORT APPROACH:

             Counting support of a set(support_count_of_sets(input_set)):                           
				counts the frequency of a candidate set                                        
             calculating multiple minimum supports:                                   
                          mis_individual():  calculates mis of every word                                    
             Sorting according to MIS:    
                           sort_accord_mis(data):  sorts the words according to the mis													 
             ms= mis_individual() (contains all the mis)
             m=sort_accord_mis(ms) (contains all the sorted 1-items according  to mis )                                                                             
             level_2_candidate_gen(m,phi):
                            generates candidate sets of size 2
				((N.B: not required as elements of f are already sorted in desired manner))
             MS_candidate_gen(f,phi):
                            generates candidate sets of size >2
             MS_apriori(data_clean,k,phi):
			Takes input data,k(the max size of frequent set)and phi(gives upper bound to the difference between min and maximum supportof elements a frequentt item set)

Otherwise,same as apriori of fixed minimum support except for

          if (  support_count_of_sets(item) /no_of_transactions)>=min_sup:
                       f.append(item)   

is replaced by

	 if(  support_count_of_sets(item)/no_of_transactions)>=ms[min(item)-1]:                           
     
                       f.append(item)      

where ms[min(item)-1] stores the mis of min elemnet of Ck-1

OUTPUT: output stored at: https://drive.google.com/drive/folders/1uoJ463rpRsqNUZAcoH9r4Z-n9I8nv6fo