See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/361910689

Graph Mining Algorithm of Multiple Objects Tracking and Data Clustering

Poster · June 2022

CITATIONS READS

0 72

2 authors:

Mustafa Asaad Hasan Tawfiq A Alasadi

University of Thi-Qar University of Babylon
33 PUBLICATIONS 44 CITATIONS 17 PUBLICATIONS 86 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Mustafa Asaad Hasan on 11 July 2022.

The user has requested enhancement of the downloaded file.

 Graph Mining Algorithm of Multiple Objects Tracking and Data Clustering
Mustafa Asaad Hasan Alkhafaji, Tawfiq A. Al-Assadi
College of Information Technology University of Babylon, Babil, Iraq
University of Thi-Qar, Nassiriyah, Iraq

Introduction
Methodology Continues of Results
Adaptive _ Graph _ Mining _ Algorithm

Graph mining is collection set of approaches and functions used to (a) properties analysis of real-world graphs, (b) guess Input: Objects Trajectory (Positions and Features)
how the properties and structure of a given graph might influence some application, and (c) evolve models that can create Output: Flatten Vectors of Fuzzy Values for each Object Trajectory
realistic graphs that match the patterns found in real-world graphs.
Step 1: Read All Trajectories as Positions and Features
a. Apriori-based approach: The priority-based chart mining (AGM) algorithm can effectively obtain all the frequently
Step 2: Apply Cosine Similarity Function between each Two Nodes for all Trajectories to Create Fuzzy Adjacency Matrixes

occurring sub-graph patterns in the chart data structure. This method is based on the breadth first search (BFS) concept,
Step 3: For Each Object Trajectory, It Will Generate and Draw Weighted Undirected Graph
which is an algorithm for searching graph data structures. It starts from the root of the tree (or some random node on the
Step 4: Apply Data Reduction on Fuzzy Adjacency Matrix by Taking Either Lower or Upper triangle of FAM
graph) and explores all adjacent vertices of the current level before moving to the next level. For example, suppose we
Step 5: Resize the Remaining Lower or Upper FAM Depend on the Minimum Number of length Vector of Object Fuzzy
have two sets of frequent elements of size-3 (abc) and (bcd). The frequent element set candidates of size generated by
Adjacency Matrix
them are simple (abcd) and are derived from the combination, and the candidate size-4 generation problem in frequent
Step 6: Convert Each FAM to Flatten Vector of the Same Size to be as Input for Clustering Algorithm
substructure mining is more complicated than the candidate generation problem in frequent element set mining, because
End
the two substructures are connected There are many ways to do it. Priority-based algorithms, such as AGM, FSG, and

path binding methods, are recently used to frequently extract substructures.

a. Pattern-Growth Approach: The priority-based method must use the strategy (BFS) because it is a step-by-step
Results Conclusions
candidate build. To identify whether the (k + 1) size chart is frequent, you need to check all its corresponding k-size Short and precise findings It is proposed a new and novel an adaptive graph mining
subgraphs to get the upper limit of its frequency. Therefore, before extracting subgraphs of any size (k + 1), the Apriori-
algorithm to mine important interested data. It is built based
based method must generally complete the extraction of subgraphs of size k. On the other hand, the pattern growth
on on some geometrical features like velocity, orientation,
method uses deep search (DFS) and its search technology is more flexible. The DFS search algorithm starts from the root
and distance of object. The data that mined by this
node or any randomly selected node, and scans as much as possible along each branch before entering other levels of
algorithm could use to represent the behaviors of any
depth. The frequent pattern growth method is a method of discovering frequent patterns and does not generate new
moving object in the scene .This graph mining algorithm is
candidate patterns. Build a FP tree instead of using Apriori's build and test strategy. The focus of the FP growth
prepared data as set of flatten vectors to fuzzy clustering
algorithm is to separate the paths of elements and extract frequent patterns. Examples of pattern growth methods are
algorithm such as c-means. When algorithm is evaluated,
gSpan, ADIMine, and DPMin. Indeed, we should use one of graph mining algorithms to clustering the trajectories by
the result showed are very high-level performance and
developing it to enhance the performance or using two or more algorithms to improve the efficiency of clustering
accuracy of data extraction from huge data representation.
process.
Acknowledgements
To my Family,,,
Hypothesis and Aims To my dear friends…
To all people that love the innovation in this world….
Create new and novel an adaptive graph mining
algorithm to mine important information from
huge data that is represented in graph and use References
[1] Israa H, Saad T. (2014). Objects Clustering of Movie Using Graph Mining
these interested info as a distinguish data for Technique.
clustering methods. [2] Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look
once: Unified, real-time object detection. In Proceedings of the IEEE conference
on computer vision and pattern recognition (pp. 779-788).
[3] Rehman, S. U., Khan, A. U., & Fong, S. (2012, August). Graph mining: A
survey of graph mining techniques. In Seventh International Conference on
Digital Information Management (ICDIM 2012) (pp. 88-92). IEEE.
View publication stats