عنوان البحث(Papers / Research Title)
Objects Clustering of Movie Using Graph Mining Technique
الناشر \ المحرر \ الكاتب (Author / Editor / Publisher)
اسراء هادي علي الشمري
Citation Information
اسراء,هادي,علي,الشمري ,Objects Clustering of Movie Using Graph Mining Technique , Time 07/07/2014 08:15:05 : كلية تكنولوجيا المعلومات
وصف الابستركت (Abstract)
Objects Clustering of Movie Using Graph Mining Technique
الوصف الكامل (Full Abstract)
Objects Clustering of Movie Using Graph Mining
Technique
Ph.D. Student Israa Hadi, Assist. Prof. Dr. Saad Talib Hasson
University of Babylon- Iraq
Abstract:-
This paper aims to develop graph mining techniques to discover patterns consisting of relationships between objects in movie film. An algorithm was designed to operate over graph data. Our approach includes a mechanism for many steps, first step discrete the movie film into number of frames (still image), second step apply segmentation technique to determine the objects in each frame and extract the features for each objects, third step construct data base consist row for each feature, fourth step construct a graph structure represent each frame, and fifth step apply adaptive graph mining algorithm to cluster the objects and determine the behavior of that objects.
Keywords:- Graph mining, data mining, Movie Graph mining.
Introduction:-
Data mining techniques can be classified according to different views such as what kinds of knowledge to work on, what kinds of databases to be mined, and what kinds of algorithms to be applied [14]. Graphs are a natural way to represent multi-relational data and are extensively used to model a variety of application domains in diverse fields. Often, in such graphs, certain sub graphs are known to possess some distinct properties and graph patterns in the proximity of these sub graphs can be an indicator of these properties. As a result, a number of approaches have been developed which have achieved promising results in uncovering interesting patterns in biological networks, social networks [1] and the World Wide Web [2]. A concise overview of the current work in in the field can be achieved by categorizing it according to the types of tasks, the types of graph data and the approaches in graph-based data mining. Much of data mining research is focused on algorithms that can discover number of features that discriminate data entities into classes or clusters. Our work is focused on data mining techniques to discover relationships between entities and determine its behaviors. Clustering, in data mining, is useful for discovering groups and identifying interesting distributions and behavior in the underlying data. The problem of clustering can be illustrated as partition the given N data points into k clusters such that the data points within a cluster are more similar to each other than data points in different clusters. Graph-based data mining (GDM) is the technique of finding new, useful, and understandable graph-theoretic patterns in a graph representation of data. There are two approaches to graph based data mining to frequent sub graph mining and graph-based relational learning. The are number of attempts in the graph mining like the AGM (Apriori Graph Mining) system which uses the Apriori level-wise approach [3]. FSG takes a similar techniques and further optimizes the algorithm for improved running times[4]. gFSG is a variant of FSG which enumerates all geometric sub graphs from the database[5]. gSpanuses DFS codes for canonical labeling and is much more memory and computationally efficient than the previous approaches[6]. FFSM is a graph mining system which uses an algebraic graph framework to address the underlying problem of sub graph isomorphism [7]. The Subdue system is a structural discovery tool that finds substructures in a graph-based representation of structural databases. This technique operates by evaluating potential substructures for their ability to compress the entire graph [8]. Subdue uses MDL-based compression heuristics, and has been applied to learning predictive as well as descriptive models. While learning descriptive models, Subdue can deal with both the single graph as well as the graph transactions category. While learning predictive models, Subdue mainly deals with the graph transactions category. GBI uses a graph size-based heuristic; the graph size definition depends on the size of the extracted patterns and the size of the compressed graph.
Dear visitor,
For downloading the full version of the research/article click on the pdf icon above.
تحميل الملف المرفق Download Attached File
|
|