Frequent pattern growth algorithm for maximizing display items

Products are goods that are available and provided in stores for sale. Products provided in stores must be arranged properly to order to attract the attention of consumers to buy. Products arranged in a store will depend on the type of store. The product arrangement at a retail store will be different from the product arrangement at a clothing store. Store display will reflect a picture that is in the store so consumers know the types of products sold by product arrangement. An attractive arrangement will stimulate the desire of consumers to buy. In data mining there are several types of methods by use including prediction, association, classification and estimation. In the prediction method there are several techniques including the frequent pattern growth (FP-growth) method. FP-growth algorithm is the development of the apriori algorithm. So, the shortcomings of the apriori algorithm are corrected by the FP-growth algorithm. FP-growth is one alternative algorithm that can be used to determine the set of data that most often appears (frequent itemset) in a data set. Results of research on the application of the FP-growth algorithm to maximizing the display of goods. It is hoped that this research can be used to adjust the product layout according to the level of frequency the product is sought by the customer so that the customer has no difficulty finding the product they want.


INTRODUCTION
Products are goods that are available and provided in stores for sale [1]. Products provided in stores must be arranged properly to attract the attention of consumers to buy. Products arranged in a store will depend on the type of store [2]. The arrangement of the product is one thing that is not less important, because this is the first impression of the visitor of the store, therefore merchandise displayed in the storeroom must be arranged so that it looks neat, harmonious and attractive to everyone, especially prospective buyers, for the arrangement of goods needed special expertise, the arrangement of goods should be changed at any time so it is not boring and adapted to the situation. Data mining is mining or discovering new information by looking for certain patterns or rules from a large amount of data that is expected to overcome these conditions [3]. Data mining is a branch of science from artificial intelligence [4]. In data mining there are several types of methods TELKOMNIKA Telecommun Comput El Control  Frequent pattern growth algorithm for maximizing display items (Asyahri Hadi Nasyuha) 391 by the use including prediction, association, classification and estimation [5]. In the prediction method there are several techniques including the frequent pattern growth (FP-growth) method. FP-growth algorithm is the development of the apriori algorithm [6]. So, the shortcomings of the apriori algorithm are corrected by the FP-growth algorithm [7]. FP-growth is one alternative algorithm that can be used to determine the set of data that most often appears (frequent itemset) in a data set [8].

Data mining
The term data mining has several views, such as knowledge discovery or pattern recognition [9]. Both of these terms have their respective accuracy. The term knowledge discovery is appropriate because the main purpose of data mining is to get knowledge that is still hidden in chunks of data [10]. The term pattern recognition is also appropriate for use because the knowledge to be extracted does indeed take the form of patterns that may also still need to be extracted from inside the chunk of data being faced.

Phase of data mining
Knowledge discovery in database (KDD) is the process of determining information that serves to determine the patterns contained in data [11]. This information is contained in a large database that was previously unknown and potentially useful. Data mining is one step in a series of KDD iterative processes [12]. The stages of the KDD process consist of [13][14][15]: data selection, pre-processing and cleaning data, transformation, data mining, and interpretation/evaluation. Data mining is one of the KDD series of knowledge. KDD is the process of determining useful information and patterns in data. Data mining is one step of an iterative KDD process. KDD stages process as shown in the following Figure 1.

Data mining techniques
Some data mining techniques are as follows [16][17][18]: − Classification: Assigning a new data record to one of some pre-defined categories (or classes). − Regression: Predict the value of a given continuous variable based on the value of another variable, assuming a linear or nonlinear dependency model. This technique is widely studied in statistics, the field of artificial neural networks (neural networks). − Clustering: Partition data set into several sub-sets or groups in such a way so that elements of a particular group have shared property sets together, with a high level of similarity in one group and a low level of similarity between groups. Also called 'unsupervised lerning'. − Association rules: Detects a collection of attributes that appear together (co-occur) on a frequent frequency, and form a number of rules of the collection. − Search for sequential patterns (sequence mining): Look for a number of events that generally occur together.
If given a set of objects, with each object associated with the time of occurrence, then get a pattern that predicts strong sequential dependencies between different events.

FP-growth
FP-growth algorithm is the development of apriori algorithm [19]. So, the shortcomings of the apriori algorithm are corrected by the FP-growth algorithm. FP-growth is one alternative algorithm that can be used to determine the set of data that most often appears (frequent itemset) in a data set [20]. Apriori algorithm requires generating candidates to get frequent itemset. However, the FP-growth algorithm generates candidates not done because FP-growth uses the concept of tree development in searching for frequent items [21]. This is what causes the FP-growth algorithm to be faster than the apriori algorithm [22].
The characteristic of FP-growth algorithm is the data structure used is a tree called FP-tree [23]. By using FP-tree, the FP-growth algorithm can directly extract frequent Itemset from FP-tree [24]. Extracting frequent itemset using the FP-growth algorithm will be done by generating a data tree structure or called FP-tree. After the FP-tree development stage of a set of transaction data, the FP-growth algorithm will be applied to look for significant frequent itemset [25].
The FP-growth algorithm is divided into three main steps [26]: 1) Conditional pattern base; 2) Conditional FP-tree; 3) Frequent itemset. The form of the FP-growth algorithm is as follows: Input: FP-tree tree Output: Rt A complete set of frequent patterns Method: FP-growth (tree, null) Procedure: FP-growth (tree, α) { 01: If Tree contains single path P; 02: Then for each combination (notated β) of the nodes in the do path 03: Generate the Build pattern β α with support from nodes in the do β path 04: Else for each a1 in the header of the do tree { 05: generate pattern 06: wake up β = a1 α with support = a1 support 07: If tree β

ANALYSIS AND RESULT 3.1. Problem analysis
Problem analysis is a process that involves a survey of the current system and analysis of user needs. After conducting interviews and reviewing documents. The next stage is to describe the system that will be designed in the form of flow, as well as unified modeling language (UML). From the results of this description can be seen what is needed for the development of the system so that the system is designed to run perfectly as desired.

System algorithm
FP-growth is a frequent itemset search algorithm that is obtained from the FP-tree by exploring the tree from bottom to top. FP-growth is the development of apriori algorithm. This algorithm determines the frequent itemset that ends in a particular suffix by using the devide and conquer method to break the problem into smaller subproblems.
In the formation of FP-growth, namely through the following algorithm: − Formation of FP-tree.
a. Determine transaction data. b. Count the amount per item. c. Determine items that meet the minimum support value≥ 20%. d. Determine transaction data that contains minimum support. − Look for frequent itemset from FP-tree. − Determine the association rules of the minimum support value and the expected value of confidence.

FP-tree formation
FP-tree is a data storage structure that is utilized. FP-tree is built by mapping each transaction data into every specific path in the FP-tree, because in every mapped transaction, there may be transactions that have the same item, so the paths are possible to overwrite each other [8]. At this stage several steps will be carried out as follows: − Pre-Processing the transaction database. To get test results, several transaction items will be tested and will be used as a transaction database. Transaction data are listed in Table 1. − Calculating the amount per item, in a number of the data Table 1 Table 1, the frequency and support values of each item are ranked from the highest based on the formula. The following data is attached in Table 2. Based on the support count value of 20%, the items used are items that have a frequency of ≥ 20%, which can be seen in Table 3. After finding the minimum support, the transaction data is moved or arranged to meet the minimum support. The data is sorted based on the highest frequency value as shown in Table 4.

Look for the frequent itemset from FP-tree
After forming the FP-tree, then look for frequent itemset, which is summarized in the Table 5. Based on the 11 frequent itemsets that have been formed above, all will be counted in the next process because they meet the frequent itemsets requirement in producing an association rule that is a minimum of 2 items if opening category, A will open category B, then there are 9 subsets that are eligible to be calculated confidence level his. After forming the FP-tree, then we are looking for the frequent itemset, which is summarized in the Table 6. After obtaining subsets that meet the requirements, then the confidence value is calculated based on a predetermined minimum confidence value of ≥ 20% to measure the validity of the association rules, as in the Table 7.   At this stage it is used to determine the value of support and confidence for each itemset using the formula: x 100 =15% x 100 =75%

Formation of association rules
Association rule is a method that aims to find patterns that often appear between many transactions, where each transaction consists of several items so that this method will contain a recommendation system through finding patterns between items in frequent transactions. Only combinations greater than or equal to the minimum confidence will be used to form a rule, the rule can be seen in Table 8.

CONCLUSION
FP-growth is a method that can process transaction data more quickly and accurately. This method can also be used to analyze sales data by determining the types of products and transactions, designing sales data grouping systems. This method can be used to arrange product appearance in order to attract customers and increase sales.