3.2.1 Working Of Prime Graph For Elementary Transaction The count of the edges which directed from P towards PMV’s or the out degree of a P is the total frequency of the appearance of P in complete set of transaction. It is shown below with the help of a figure.
3.2.2 Working Of Prime Graph For Subset Transaction The count of the edges which directed from one PMV towards other PMV or the same PMV is the total frequency of the appearance of particular subset of itemset in a complete set of transaction.
It is shown below with the help of figure. Figure 3.3: PC-Graph Construction For Subset Transaction. 3.3 Prime_Graph Miner AlgorithmAs explained in previous section, during construction of Prime graph following registers maintained by storing the updated valuea) count- which stores the frequency of particular items.b) Local-count- Keeps the value of current PMV.
c) Global-counting-keep track on the frequency of frequent and infrequent items.d) StatusStep1. Traverses the graph in top-down directionStep2. Compare the frequency of the elementary itemset (pages) to the user defined thresholdStep3. Prunes the infrequent itemsetsStep4.
Matches the subset of the transactions with one another with the help of PMVStep5. Compares the frequency of repeated subset transaction with the user defined threshold.Step6.
Results gives the frequent elementary itemset of the frequent page numbers and the frequent subsets of the transactions that are same set of pages repeated in more than one transaction. The Prime_Graph Miner algorithm traverses the completed Prime Graph to discover the MFP in top-down direction. There is no need to database scan again, because all information about items and patterns are stored in the Prime Graph itself.
The miner algorithm prunes the infrequent itemsets. As a result the search space is reduced, which dramatically reduces the computation and memory requirement and enhances the efficiency. Table 2 shows the item frequency and considered prime number for transaction database TABLE 2 Frequency of items and supposed prime numbersPage no. Prime number Item Frequency8 2 65 3 311 5 720 7 76 11 39 13 44 Experimental Results All experiments were performed in a time-sharing environment in a 2.
4 GHz PC. All the algorithms are implemented using Matlab. In first experiment we use synthetic web log sparse datasets .
The number of transactions are 50, the average transaction Login is 12 and the number of transaction increased from 50 to 100 to evaluate how the data transformation technique can compact the size of dataset. Fig.4.
1 shows comparison of the size of original dataset with the size of transformed dataset using proposed data transformation technique.Figure 4.1 Size comparison of data Second experiment is to compare the performance of the Prime Graph with the FP-Tree on the web log dataset. To allow a fair comparison of algorithms, firstly plot all transactions using Prime Graph and FP-Tree separately. Time taken by six random sets of 50 transactions of 12 logins are recorded to plot a comparative graph between tree construction and prime-graph. Then all frequent patterns are generated by same procedure run in cached mode.
Fig. 4.4 shows the PC_Graph outperforms the performance of FP-Tree.Hence, this is proved by the experiments that proposed method is an improvement of previous method that has been proposed for maximal frequent pattern mining and it requires only one database scan. The experimental result verifies the compactness and the efficiency of Prime Graph method. Figure 4.4 comparative analyses of FP-Tree and prime-Graph5 Conclusion and Future workThis Proposed method concludes that Prime graph method is a technique based on without candidate generation so it does not produce large number of frequent candidates to generate further frequent patterns. It requires only one database scan to mine the frequent patterns as all the useful information about the transaction stores in the Prime graph itself.
It requires less search space as Miner algorithm Prunes the infrequent itemsets by using combination of subset and superset pruning techniques which reduces the size of dataset up to an extent. It is time efficient; as time required in constructing FP-Tree is much greater than the time needed to plot a Prime graph with the same set of data.This method is an improvement over FP-Tree method in terms of time, space and speed. This method has an advantage over other methods that it is independent of size of dataset, whatever be the size of transaction it can be transformed into prime number and it gives frequency of both elementary itemsets as well as subsets 2. Our proposed method, is simple to implement, easy to understand and does not includes any complex structures.
This graphical method can extended up to wide applications for enhancing performance of the particular like It can be applied in incremental mining of frequent patterns where database transactions are updated regularly. In addition, it can also be used for interactive mining of frequent patterns where minimum support threshold can be changed to find new correlation between patterns. This method can be used for large graph structures with unique nodes and can be applied to large databases to find out the particular subset repetition of the transaction which can be useful to avoid frauds as well as can be useful in discovering knowledge for artificial intelligence based applications.