Mining frequent itemsets using advanced partition approach

MINING FREQUENT ITEMSETS USING ADVANCED PARTITION APPROACH Nay Chi Lynn, Khin Myat Myat Moe University of Computer Studies, Magway naychelynn@gmail.com promotion, etc. Market basket analysis studies the buying habits of customers by finding frequent itemsets between the different items that customers purchase [3]. Abstract Frequent itemsets mining plays an important part in many data mining tasks. This technique has been used in numerous practical applications, including market basket analysis. This paper presents mining frequent itemsets in large database of medical sales transaction by using the advanced partition approach. This advanced partition approach executes in two phases. In phase 1, the advanced partition approach logically divides the database into a number of non-overlapping partitions. These partitions are considered one at a time and all local frequent itemsets for those partitions are generated using the apriori method. In phase 2, the advanced partition approach finds the final set of frequent itemsets. The purpose of this paper is to extract the final sets of frequent itemsets from medical retail datasets and to support efficient information used to plan marketing or advertising strategies for medical stores and companies. Algorithms for finding frequent itemsets like Apriori, needs many database scans. But, this advanced partition approach needs to scan the entire database only one time. So, it reduces the time taken for the large database scan in mining frequent itemsets. Mining frequent itemsets is important and interesting to the fundamental research in the mining of association rules. Frequent itemsets mining, one technique of the descriptive mining, is widely used for market basket analysis. It analyses customer buying habits by finding frequent itemsets between different items that occur frequently together in a given set of data. There are many algorithms for finding frequent itemsets. The Apriori is the basic first algorithm for finding frequent itemsets. In this system, the advanced partition algorithm is used for finding frequent itemsets with an entire single data pass. By the result of system testing, the user can know which algorithm is suitable for his medical company and store. 2. Related work The works related to this system are presented in here. Frequent itemsets mining is well explored for various data types. The Apriori algorithm also called level-wise algorithm was proposed by Agrawal and Srikanth in 1994. The name of the algorithm is based on the fact that the algorithm uses the prior knowledge of frequent itemsets properties [1]. Nguyen and Orlowska show the data partition approach to further improve the performance of frequent itemsets computation. The methods focus on potential reduction of the size of the input data required for deployment of the partitioning based algorithms [4]. Kranthi and Malreddy have proposed the advanced partition 1. Introduction Nowadays, retailing is becoming a highperformance sport. Like athletes, retailers are becoming competitive, seeking technology to gain, and trying to have more knowledge into customer buying behavior. Market basket analysis has emerged as a step of the retail merchandising, 1 partition for calculating the frequent itemsets separately the local minimum support is set to 1. At the end of phase I, the advanced partition approach merges all local frequent itemsets of each partition to generate the global candidate item sets (Ck G ). Phase II: This phase just prune the item sets from the global candidate itemsets list whose combined support (s(c) Tc) (total support of an item set in all the partitions) is less than the global minimum support. So here the advanced partition approach reads the entire database once during the Phase I. And also, partition sizes are chosen such that each partition can be accommodated in the main memory [2]. approach to generate the frequent itemsets in a single pass over the database [2]. In this paper the advanced partition approach has been used to find the frequent itemsets. 3. Theory background Frequent itemsets mining is an important data mining task. It extracts interesting correlations, frequent itemsets among sets of items in the transaction databases. 3.1 Frequent Itemsets A set of items is referred to as an itemset. An item set that contains k items is a k- itemset .The set {computer, finical-management software} is a 2itemset. The occurrence frequency of an itemset is the number of transactions that contain the itemset. An itemset satisfies minimum support if the occurrence frequency of the itemset is greater than or equal to the product of min-sup and the total number of transactions in the database. The number of transactions required for the itemset to satisfy minimum support is therefore referred to as the minimum support count. If an itemset satisfies minimum support, then it is a frequent itemset [1]. Table 1 Notations used for advanced partition approach Notation Li CkG Lik LG 3.2 Advanced Partition Approach s(c)Tc The advanced partition approach is based on the premise that number of items in a transaction is quite less compared to total number of items in the transaction database. This advanced partition approach is efficient when the local support of each frequent item set in a partition is much higher than 1. The advanced partition approach executes in two phases. In phase 1, the advanced partition approach divides the database into a number of non-overlapping partitions. These partitions are considered one at a time and all frequent itemsets for that partition (Li) are generated using the Apriori algorithm (presented in section 3.2.1). In addition, when taking each Meaning Local Frequent Sets: Set of Local Frequent Itemsets of partition i. Global Frequent Sets: Set of global candidate k-itemsets. Local Frequent Sets: Set of local frequent k-Itemsets in partition i. Global Frequent Sets: Set of global frequent Itemsets. Combined Support Total support of candidate set c in all partitions. Below is the algorithm of advanced partition approach: P = partition_database (T); N = Number of partitions; // Phase I for i= 1 to n do begin read_in_partition ( Ti in P ) Li = generate all frequent itemsets of Ti using Apriori algorithm in main memory. end // Merge Phase for (k = 2; L i k ≠ ф, i=1,2,…,n; i++) do begin 2 CkG = Ү ni=1 L ik end // Phase II LG = ф ; for each c є C G do begin if s(c) TC ≥ σ LG = LG U {s(c)} end Answer = LG 5. Ct = subset (Ck, t); // get the subsets of t that are candidates 6. for each candidate c є Ct 7. c.count++; 8. } 9. Lk = { c є Ck\c.count ≥ min_sup} 10. } 11. return L = UkLk; Procedure apriori_gen (Lk-1: frequent (k – 1)itemsets; min_sup: minimum support threshold) 1. for each itemset l1є Lk-1 2. for each itemset l2 є Lk-1 3. if(l1[1] = l2[1]) ^(l1[2] = l2[2]) ^ ... ^ (l1[k – 2] = l2[k – 2]) ^ (l1[k – 1] < l2[k – 1]) then { 4. c = l1 ∞ l2; // join step: generate candidates 5. if has_infrequent_subset(c, Lk-1) then 6. delete c; // prune step: remove unfruitful candidate 7. else add c to Ck; 8. } 9. return Ck; Figure 1: The Advanced partition approach for discovering frequent itemsets. 3.2.1 Apriori Algorithm Apriori is an influential algorithm for mining frequent itemsets for Boolean association rules. Apriori uses prior knowledge of frequent item set properties. Apriori employs an iterative approach known as level-wise search, where k itemsets are used to explore (k+1)-item sets [1]. There are two-steps in Apriori Algorithm: 1. The join step: To find Lk, a set of candidate k- itemsets is generated by joining Lk-1 with itself. This set of candidates is denoted Ck. Let l1 and l2 be itemsets in Lk-1. 2. The prune step: Ck is a superset of Lk, that is, its members may or may not be frequent, but all of the frequent k-itemsets are included in Ck. A scan of the database to determine the count of each candidate in Ck would result in the determination of Lk. Algorithm: Apriori. Find frequent itemsets using an interactive level-wise approach based on candidate generation. Input: Database, D, of transactions; minimum support threshold, min_sup. Output: L, frequent itemsets in D. Method: 1. 2. 3. 4. Procedure has_infrequent_subset (c: candidate kitemset; Lk-1: frequent (k – 1)-itemsets); // use prior knowledge 1. for each (k – 1)-subset s of c 2. if s ¢ Lk-1 then 3. return TRUE; return FALSE;[1] Figure 2: The Apriori algorithm 4. System design and implementation This paper finds frequent itemsets between medical products purchased together by customers. For the purpose of implementing, the retail datasets supplied by a medical Co, Ltd is used. Using advanced partition approach, the system is efficient in computing frequent itemsets and can reduce the time spent in performing the I/O operations for large databases. L1= find_frequent_1-itemsets(D); for (k = 2; Lk-1 ≠ ф; k++) { Ck = apriori_gen (Lk-1, min_sup); for each transaction t є D { // scan D for counts 3 4.1 4.2 Implementation of the advanced partition approach System Design If the user wants to entry new medical products into his data set, the Entry new items process should be chosen. If the user wants to buy those above medical products, the Sales process should be chosen. Moreover, the user can also import the items from other datasets using import data menu. The system will find the global frequent itemsets by using the advanced partition approach. The user has to choose the global minimum support count either number or percentage. Below is the example which shows the working of the system. The transaction dataset (T) consists of a set of transactions in the form (tid, itemspurchased) where tid is transaction ID. This system includes medical sales items such as, Lensan Para, Para BPI, Amoxy, I Amox, Flumox, Brumox, Biogesic, etc. The advanced partition approach executes in two phases. In this form, medical sales itemsets are partitioned according to user specified input partition number (N). If the user input partition number (N) is 2, then the advanced partition approach (phase 1) divides sales itemsets into two partitions such as partition 1 and partition 2 as shown in figure 4. begin Choose process menu Entry new items Sales process Calculate frequent itemsets using advanced partition approach Import data Transactio n data Choose (global) frequent itemsets s(c)>=(global) min-sup no Prune those itemsets yes Generate global frequent itemsets Display global frequent itemsets end Figure 4: Partition dataset form Figure 3: System flow diagram Then, the advanced partition approach (phase 1) finds all local frequent itemsets of each partition using Apriori method in figure 5. 4 addressed in mining frequent itemsets for discovering the set of large items. This system is intended to implement frequent itemsets mining. The discovery of frequent patterns and correlation relationships among huge amounts of data is useful in selective marketing, decision analysis, and business management. This system can help retailers, buyers, planners, merchandisers, and store managers to plan more profitable advertising and promotions, attract more customers, and increase the value of the market basket. Moreover, one can use the results to plan marketing or advertising strategies, or in the design of a new catalog. For instance, it may help managers to design different store layouts. In one strategy, items that are frequently purchased together can be placed together in close proximity in order to further encourage the sales of such items together. This system can act as a consultant for medical stores by giving the information of frequent items. Figure 5: Local frequent itemsets form The advanced partition approach (phase 2) finds global frequent itemsets which satisfy userspecified minimum support (if user specified minimum support count number is 10, it will show the itemsets above 10) in figure 6 and then gives final set of (global) frequent itemsets. References [1] J.W.Han, M. Kamber, “Data Mining Concepts and Technique”, ISBN 1-55860-489-8, Morgan Kaufmann Publishers. [2] Kranthi K. Malreddy, B.S, “Mining Frequent Itemsets Using Advanced Partition Approach”, Dean of the Graduate School, December, 2004. [3] Lary Gordan, Partner, “Leading Practices In Market Basket Analysis”, the Face Point Group, 349 First Street, Los Altos, CA 94022 (650)559-2105, Gordan@Facepoint.Com [4] Son N. Nguyen, Maria E. Orlowska, “A Further Study In The Data Partitioning Approach For Frequent Itemsets Mining”, School Of Information Technology And Electrical Engineering, the University Of Queensland, QLD 4072, Australia {Nnson, Maria} Itee.Uq.Edu.Au. Figure 6: Final set of frequent itemsets form 5. Conclusions Mining frequent itemsets is important and is one of the primary sub-areas on the fields of data mining. Market basket data analysis has been well 5