• The KIPS Transactions:PartD
• /
• v.12D no.6 s.102
• /
• pp.817-828
• /
• 2005

The main research problems in a mining frequent itemsets are reducing memory usage and processing time of the mining process, and most of the previous algorithms for finding frequent itemsets are based on an Apriori-property, and they are multi-scan algorithms. Moreover, their processing time are greatly increased as the length of a maximal frequent itemset. To overcome this drawback, another approaches had been actively proposed in previous researches to reduce the processing time. However, they are not efficient on a sparse .data set This paper proposed an efficient mining algorithm for finding frequent itemsets. A novel tree structure, called an $L_2$-tree, was proposed int, and an efficient mining algorithm of frequent itemsets using $L_2$-tree, called an $L_2$-traverse algorithm was also proposed. An $L_2$-tree is constructed from $L_2$, i.e., a set of frequent itemsets of size 2, and an $L_2$-traverse algorithm can find its mining result in a short time by traversing the $L_2$-tree once. To reduce the processing more, this paper also proposed an optimized algorithm $C_3$-traverse, which removes previously an itemset in $L_2$ not to be a frequent itemsets of size 3. Through various experiments, it was verified that the proposed algorithms were efficient in a sparse data set.

• Journal of KIISE:Databases
• /
• v.36 no.5
• /
• pp.341-351
• /
• 2009

Recently high utility pattern (HUP) mining is one of the most important research issuer in data mining since it can consider the different weight Haloes of items. However, existing mining algorithms suffer from the performance degradation because it cannot easily apply Apriori-principle for pattern mining. In this paper, we introduce new high utility pattern mining approach by using a prefix-tree as in FP-Growth algorithm. Our approach stores the weight value of each item into a node and utilizes them for pruning unnecessary patterns. We compare the performance characteristics of three different prefix-tree structures. By thorough experimentation, we also prove that our approach can give performance improvement to a degree.

• Journal of Digital Convergence
• /
• v.13 no.9
• /
• pp.269-276
• /
• 2015

Recently, with the activation of the industry related to the big data, the global security companies have expanded their scopes from structured to unstructured data for the intelligent security threat monitoring and prevention, and they show the trend to utilize the technique of user's tendency analysis for security prevention. This is because the information scope that can be deducted from the existing structured data(Quantify existing available data) analysis is limited. This study is to utilize the analysis of security tendency(Items classified purpose distinction, positive, negative judgment, key analysis of keyword relevance) applying the machine learning algorithm($Na{\ddot{i}}ve$ Bayes, Decision Tree, K-nearest neighbor, Apriori) in the big data environment. Upon the capability analysis, it was confirmed that the security items and specific indexes for the decision of security tendency could be extracted from structured and unstructured data.

• The KIPS Transactions:PartD
• /
• v.13D no.5 s.108
• /
• pp.651-660
• /
• 2006

Algorithms for mining association rules based on the Apriori algorithm use the hash tree data structure for storing and counting supports of the candidate frequent itemsets and the most part of the execution time is consumed for searching in the hash tree. The DHP(Direct Hashing and Pruning) algorithm makes efforts to reduce the number of the candidate frequent itemsets to save searching time in the hash tree. For this purpose, the DHP algorithm does preparative simple counting supports of the candidate frequent itemsets. At this time, the DHP algorithm uses the direct hash table to reduce the overhead of the preparative counting supports. This paper proposes and evaluates an efficient hashing mechanism for the direct hash table $H_2$ which is for pruning in phase 2 and the hash tree $C_k$, which is for counting supports of the candidate frequent itemsets in all phases. The results showed that the performance improvement due to the proposed hashing mechanism was 82.2% on the maximum and 18.5% on the average compared to the conventional method using a simple mod operation.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.22 no.50
• /
• pp.183-193
• /
• 1999

The problem of discovering association rules has received considerable research attention and several fast algorithms for mining association rules have been developed. In this paper, we propose an efficient algorithm for mining quantitative association rules with item constraints. For categorical attributes, we map the values of the attribute to a set of consecutive integers. For quantitative attributes, we can partition the attribute into values or ranges. While such constraints can be applied as a post-processing step, integrating them into the mining algorithm can reduce the execution time. We consider the problem of integrating constraints that are boolean expressions over the presence or absence of items containing quantitative attributes into the association discovery algorithm using Apriori concept.

• Journal of Intelligence and Information Systems
• /
• v.14 no.1
• /
• pp.1-17
• /
• 2008

Data mining is defined as the process of discovering meaningful and useful pattern in large volumes of data. In particular, finding associations rules between items in a database of customer transactions has become an important thing. Some data structures and algorithms had been proposed for storing meaningful information compressed from an original database to find frequent itemsets since Apriori algorithm. Though existing method find all association rules, we must have a lot of process to analyze association rules because there are too many rules. In this paper, we propose a new data structure, called a Frequent Pattern Network (FPN), which represents items as vertices and 2-itemsets as edges of the network. In order to utilize FPN, We constitute FPN using item's frequency. And then we use a clustering method to group the vertices on the network into clusters so that the intracluster similarity is maximized and the intercluster similarity is minimized. We generate association rules based on clusters. Our experiments showed accuracy of clustering items on the network using confidence, correlation and edge weight similarity methods. And We generated association rules using clusters and compare traditional and our method. From the results, the confidence similarity had a strong influence than others on the frequent pattern network. And FPN had a flexibility to minimum support value.

• Journal of Internet Computing and Services
• /
• v.16 no.1
• /
• pp.67-74
• /
• 2015

In recent years, frequent itemset mining for considering the importance of each item has been intensively studied as one of important issues in the data mining field. According to strategies utilizing the item importance, itemset mining approaches for discovering itemsets based on the item importance are classified as follows: weighted frequent itemset mining, frequent itemset mining using transactional weights, and utility itemset mining. In this paper, we perform empirical analysis with respect to frequent itemset mining algorithms based on transactional weights. The mining algorithms compute transactional weights by utilizing the weight for each item in large databases. In addition, these algorithms discover weighted frequent itemsets on the basis of the item frequency and weight of each transaction. Consequently, we can see the importance of a certain transaction through the database analysis because the weight for the transaction has higher value if it contains many items with high values. We not only analyze the advantages and disadvantages but also compare the performance of the most famous algorithms in the frequent itemset mining field based on the transactional weights. As a representative of the frequent itemset mining using transactional weights, WIS introduces the concept and strategies of transactional weights. In addition, there are various other state-of-the-art algorithms, WIT-FWIs, WIT-FWIs-MODIFY, and WIT-FWIs-DIFF, for extracting itemsets with the weight information. To efficiently conduct processes for mining weighted frequent itemsets, three algorithms use the special Lattice-like data structure, called WIT-tree. The algorithms do not need to an additional database scanning operation after the construction of WIT-tree is finished since each node of WIT-tree has item information such as item and transaction IDs. In particular, the traditional algorithms conduct a number of database scanning operations to mine weighted itemsets, whereas the algorithms based on WIT-tree solve the overhead problem that can occur in the mining processes by reading databases only one time. Additionally, the algorithms use the technique for generating each new itemset of length N+1 on the basis of two different itemsets of length N. To discover new weighted itemsets, WIT-FWIs performs the itemset combination processes by using the information of transactions that contain all the itemsets. WIT-FWIs-MODIFY has a unique feature decreasing operations for calculating the frequency of the new itemset. WIT-FWIs-DIFF utilizes a technique using the difference of two itemsets. To compare and analyze the performance of the algorithms in various environments, we use real datasets of two types (i.e., dense and sparse) in terms of the runtime and maximum memory usage. Moreover, a scalability test is conducted to evaluate the stability for each algorithm when the size of a database is changed. As a result, WIT-FWIs and WIT-FWIs-MODIFY show the best performance in the dense dataset, and in sparse dataset, WIT-FWI-DIFF has mining efficiency better than the other algorithms. Compared to the algorithms using WIT-tree, WIS based on the Apriori technique has the worst efficiency because it requires a large number of computations more than the others on average.

• Journal of KIISE:Software and Applications
• /
• v.29 no.1_2
• /
• pp.80-90
• /
• 2002

Clustering is a useful technique for grouping data points such that points within a single group/cluster have similar characteristics. Many clustering algorithms have been developed and used in engineering applications including pattern recognition and image processing etc. Recently, it has drawn increasing attention as one of important techniques in data mining. However, clustering algorithms such as K-means and Fuzzy C-means suffer from difficulties. Those are the needs to determine the number of clusters apriori and the clustering results depending on the initial set of clusters which fails to gain desirable results. In this paper, we propose a new clustering algorithm, which solves mentioned problems. In our method we use evolutionary algorithm to solve the local optima problem that clustering converges to an undesirable state starting with an inappropriate set of clusters. We also adopt a new measure that represents how well data are clustered. The measure is determined in terms of both intra-cluster dispersion and inter-cluster separability. Using the measure, in our method the number of clusters is automatically determined as the result of optimization process. And also, we combine heuristic that is problem-specific knowledge with a evolutionary algorithm to speed evolutionary algorithm search. We have experimented our algorithm with several sets of multi-dimensional data and it has been shown that one algorithm outperforms the existing algorithms.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• v.1
• /
• pp.277-282
• /
• 2006

It is well known that Kalman filtering is an optimal real-time data fusion method for GPS/INS integration. However, it has some limitations in terms of stability, adaptability and observability. A Kalman filter can perform optimally only when its dynamic model is correctly defined and the noise statistics for the measurement and process are completely known. It is found that estimated Kalman filter states could be influenced by several factors, including vehicle dynamic variations, filter tuning results, and environment changes, etc., which are difficult to model. Neural networks can map input-output relationships without apriori knowledge about them; hence a proper designed neural network is capable of learning and extracting these complex relationships with enough training. This paper presents a GPS/INS integrated system that combines Kalman filtering and neural network algorithms to improve navigation solutions during GPS outages. An Extended Kalman filter estimates INS measurement errors, plus position, velocity and attitude errors etc. Kalman filter states, and gives precise navigation solutions while GPS signals are available. At the same time, a multi-layer neural network is trained to map the vehicle dynamics with corresponding Kalman filter states, at the same rate of measurement update. After the output of the neural network meets a similarity threshold, it can be used to correct INS measurements when no GPS measurements are available. Selecting suitable inputs and outputs of the neural network is critical for this hybrid method. Detailed analysis unveils that some Kalman filter states are highly correlated with vehicle dynamic variations. The filter states that heavily impact system navigation solutions are selected as the neural network outputs. The principle of this hybrid method and the neural network design are presented. Field test data are processed to evaluate the performance of the proposed method.