INTERACTIVE MINING

This paper focuses on interactive Knowledge Discovery pro-cesses in the context of understanding an activity from behavioural data. Data mining provides patterns experts have to interpret and synthesize as new knowledge. Discovering patterns is an analysis task while build-ing new symbolic knowledge is a synthesis task. A previous trace based approach (Abstract) offered a first answer to support analysis. This paper goes one step forward in supporting the synthesis task. We mod-ify an algorithm of automata discovery in order to involve the user in the mining process, exploiting his expert knowledge about the observed activity. We chose the α-algorithm (Van Der Aalst et al.) developed for Petri nets discovery in a workflow management context. The modified algorithm is described and illustrated, showing how to use intermediate data to converge interactively to a satisfying automata. Finally, we dis-cuss the use of this approach to contribute to a new knowledge mining process.

Web usage mining is about analyzing the user interactions with a web server based on resources like web logs, click streams and database transactions. It helps in discovering the browsing patterns of the user and in relating the pages visited by him. This knowledge can be of help in making business decisions, refining the web site design to derive personalized pages. Web usage mining uses the data mining and the incremental mining techniques to extract the user usage patterns and study the visiting characteristics of the user. Normally, the source for all this data is the web server log, which changes dynamically. Conventional data mining techniques were proposed to be inefficient, as they need to be re-executed every time the log changes which in turn requires multiple database scans. This paper discusses an algorithm to suit the dynamically changing log scenario, based on association rule mining with incremental technique. The algorithm proved to be more efficient as there are redundant no of scans of database; it avoids the generation of candidate keys and allows interactive mining with different supports. The experimental results proving this claim are given in this paper.

This paper addresses the incremental and decremental maintenance of the frequent pattern space. We conduct an in-depth investigation on how the frequent pattern space evolves under both incremental and decremental updates. Based on the evolution analysis, a new data structure, Generator-Enumeration Tree (GE-tree), is developed to facilitate the maintenance of the frequent pattern space. With the concept of GE-tree, we propose two novel algorithms, Pattern Space Maintainer+ (PSM+) and Pattern Space Maintainer-(PSM-), for the incremental and decremental maintenance of frequent patterns. Experimental results demonstrate that the proposed algorithms, on average, outperforms the representative state-of-the-art methods by an order of magnitude.

Big-data takes us into new epoch of data, which commonly referred to be as a big data. It pose a challenges to the researchers with more velocity, more variety and large volumes and software engineers are working on variety of methods in avoiding cost of development process. Where the commonly used software’s are not able to imprisonment, accomplish and process within the lapsed time. Furthermore, there is a need to discover new procedures for to process large volumes of data to optimization, datamining and knowledge discovery. This ambitions and motivations are drives the researchers to Big-data analytics and big-data mining. Over the earlier few centuries, different procedures have been proposed to use the MapReduce model-which decrease the space of search with distributed or parallel computing-for different big data mining and analytics tasks and these methods are used in providing a cost effective solution across various software development models using Bayesian Approaches. In ...

Recently, knowledge extraction from transactional graph databases by mining frequent subgraph patterns has become an interesting research topic. One of the important challenges in this topic is the situation called interactive mining in which the minimum support threshold can be frequently changed to find proper frequent subgraph patterns. Obviously, running the mining method from scratch is very costly. Therefore, in this paper, an efficient tree called IDFP-tree is proposed to develop our previous two-layer model for interactive mining of frequent subgraph patterns from transactional graph databases. The proposed tree constructs the mining model separated from the mining process. Thus, when the minimum support threshold is changed, there is no need to reconstruct the mining model, and only the mining process must be rerun. The experimental results show when the mining model is constructed by IDFP-tree, it can be frequently used and the total runtime of interactive mining of freque...

Task of extracting fruitful knowledge from huge datasets is called data mining. It has several aspects like predictive modeling or classification, cluster analysis, association analysis, anomaly detection and regression etc. Among all association rule mining is one of the major tasks for data mining. Association analysis is mainly used to discover patterns, which describes strongly associated features in the data. Market basket data is one of the major applications of association rule mining. Other applications include bioinformatics, medical diagnosis, scientific data analysis, web mining, finding the relationships between different elements of earth climate system etc. Various algorithms have been proposed by researchers to improve the performance of frequent pattern mining such as Apriori, Frequent Pattern (FP)-growth etc. We are providing a brief description of some of the techniques in detail in the later section of this paper.

Association rule mining is a very efficient technique for finding a strong relation between correlated data. For the mining of positive and negative rules, a variety of algorithms are used such as Apriori algorithm. Practically the data set can contain a large number of data entities which form a large number of rules set. For the understanding of efficient relation, it is needed that rules should be kept as minimum as possible without losing the useful information. This is a difficult task and can be seen as optimization problem. To eliminate the multi-scan problem and reduce a large number of mined rules, in this paper, we propose a novel approach to mine interesting positive and negative rules from frequent and infrequent pattern set based on genetic Tabu heuristic search with key innovation to optimize genetic population of rules. Genetic Tabu Algorithm (GTA) follows procedures similar to the Genetic Algorithm (GA). However, for the selection process, GTA uses the Tabu search pr...

Recently, knowledge extraction from transactional graph databases by mining frequent subgraph patterns has become an interesting research topic. One of the important challenges in this topic is the situation called interactive mining in which the minimum support threshold can be frequently changed to find proper frequent subgraph patterns. Obviously, running the mining method from scratch is very costly. Therefore, in this paper, an efficient tree called IDFP-tree is proposed to develop our previous two-layer model for interactive mining of frequent subgraph patterns from transactional graph databases. The proposed tree constructs the mining model separated from the mining process. Thus, when the minimum support threshold is changed, there is no need to reconstruct the mining model, and only the mining process must be rerun. The experimental results show when the mining model is constructed by IDFP-tree, it can be frequently used and the total runtime of interactive mining of freque...

An increasing number of efficient methods have been proposed to mine frequent patterns from uncertain data obtained from real applications such as social networks and life-sciences. Since these data are constantly being updated, needs of the users are changed and they adjust a new minimum support threshold to find new and proper frequent patterns. Obviously, finding the new frequent patterns by running the algorithm from scratch is very costly especially when the database is very large. In this paper, an efficient tree called UDFP-tree is proposed for interactive mining from uncertain data. The proposed tree aims to construct the mining model separated from the mining process. The experimental results show that by using UDFP-tree, there is no need to reconstruct the mining model when user changes the minimum support threshold.

This paper essentially analyses the sequential pattern of mining algorithms. The discovery of Association relationship seeks more attention in data mining due to the constantly increasing amount of data stored in the real application system. Mining for association rules has its usage in several areas of business such as the process of decision making and the development of customized marketing programs & strategies. Therefore, the primary objective of data mining is to transform “data into knowledge”. As a result of which, mining association rules from enormous databases has been a significant topic in recent research for knowledge discovery in databases.

The market basket has been finded patterns of purchase customer in SME. Purchase patterns can help to make recommendations and product promotions. This research used K-Means algorithm for sales data clustering and uses FP-Growth Algorithm to know the relation of each cluster. K-Means clustering to classify customer data based on the same attribute, then determined the relationship between patterns in each group with FP-Growth Algorithm. K-Means to do customer segmentation based on background, customer characteristic and level of purchasing power. To facilitate the analysis of customer relationships with products purchased, then each cluster profilling customer will be processed data record by FP Growth to know the relevance of goods purchased. The research presents a discussion of the comparison of time complexity between FP-Growth algorithms and Apriori Algorithms. This research would been done the development and application of the use of Trees ie FP-Tree (Frequent Pattern Tree). They are an extension of the use of Trees in the data structure. FP-Tree is used in conjunction with the FP-Growth algorithm to determine the frequent itemset of a database, in contrast to the a priori paradigm of scanning the database repeatedly to determine the frequent itemset. In this study, the number of transactions with many items of goods and consumer purchasing power are varied, grouped first by using K-Means algorithm, cluster results formed into several groups including five customer groups based on customer profile. The result of the test is average on minsupp = 60 and minconf = 40, so the average processing time is 957 ms.

The market basket has been finded patterns of purchase customer in SME. Purchase patterns can help to make recommendations and product promotions. This research used K-Means algorithm for sales data clustering and uses FP-Growth Algorithm to know the relation of each cluster. K-Means clustering to classify customer data based on the same attribute, then determined the relationship between patterns in each group with FP-Growth Algorithm. K-Means to do customer segmentation based on background, customer characteristic and level of purchasing power. To facilitate the analysis of customer relationships with products purchased, then each cluster profilling customer will be processed data record by FP Growth to know the relevance of goods purchased. The research presents a discussion of the comparison of time complexity between FP-Growth algorithms and Apriori Algorithms. This research would been done the development and application of the use of Trees ie FP-Tree (Frequent Pattern Tree). They are an extension of the use of Trees in the data structure. FP-Tree is used in conjunction with the FP-Growth algorithm to determine the frequent itemset of a database, in contrast to the a priori paradigm of scanning the database repeatedly to determine the frequent itemset. In this study, the number of transactions with many items of goods and consumer purchasing power are varied, grouped first by using K-Means algorithm, cluster results formed into several groups including five customer groups based on customer profile. The result of the test is average on minsupp = 60 and minconf = 40, so the average processing time is 957 ms.

—Commonly, frequent patterns are mined by satisfying a user specified minimum support threshold or minsup in short. In some applications, finding proper frequent patterns by changing the value of minsup is needed. Since rerunning the mining algorithm from scratch can be very time consuming and may unacceptable for real time applications, researchers have introduced interactive mining. Although need to interactive mining has been addressed in many studies, thus far there has not been proposed any specified model to develop an efficient interactive mining method. In this paper, we propose a two-layer model including mining model construction and mining process for interactive mining. The experimental results verify that using the proposed model avoids database rescanning and reconstructing of the mining model which is the basic idea to enhance the efficiency of interactive mining.

An increasing number of efficient methods have been proposed to mine frequent patterns from uncertain data obtained from real applications such as social networks and life-sciences. Since these data are constantly being updated, needs of the users are changed and they adjust a new minimum support threshold to find new and proper frequent patterns. Obviously, finding the new frequent patterns by running the algorithm from scratch is very costly especially when the database is very large. In this paper, an efficient tree called UDFP-tree is proposed for interactive mining from uncertain data. The proposed tree aims to construct the mining model separated from the mining process. The experimental results show that by using UDFP-tree, there is no need to reconstruct the mining model when user changes the minimum support threshold.

Recently, knowledge extraction from transactional graph databases by mining frequent subgraph patterns has become an interesting research topic. One of the important challenges in this topic is the situation called interactive mining in which the minimum support threshold can be frequently changed to find proper frequent subgraph patterns. Obviously, running the mining method from scratch is very costly. Therefore, in this paper, an efficient tree called IDFP-tree is proposed to develop our previous two-layer model for interactive mining of frequent subgraph patterns from transactional graph databases. The proposed tree constructs the mining model separated from the mining process. Thus, when the minimum support threshold is changed, there is no need to reconstruct the mining model, and only the mining process must be rerun. The experimental results show when the mining model is constructed by IDFP-tree, it can be frequently used and the total runtime of interactive mining of frequent subgraph patterns can be reduced.