Graph Analytics with Data Science Languages

In Big Data analytics, data exploded in the three Vs: Volume, Velocity, and Variety. The three Vs brought new challenges to data analysis systems, which require new approaches, tools, and algorithms to analyze data. The most complex exploration mechanism in Big Data Analytics is graphs, which are flexible to represent any set of interconnected objects. Graph analytics is particularly challenging mainly due to large graph sizes and the structure of graphs. In this dissertation, we work on analyzing large graphs that cannot fit in main memory. First, we extract a general computation pattern for several graph properties that can be solved with iterative algorithms. Then, we start with database management systems (DBMSs) to compute those graph properties since a lot of data stored in DBMSs can be analyzed as graphs. We proposed algorithms and optimized query solutions with a focus on graph partitions. Experimental evaluations demonstrate that our solutions can work on large graphs with good speed up and reasonable performance. After conducting a comprehensive survey about Data Science Languages, we continued our work on analyzing large graphs with Data Science Language, Python. We developed a lightweight C++ function that can be used for several graph algorithms. The function is easily called in Python. Comparing our function with other state-of-the-art graph libraries shows its good performance. Finally, we study the crucial graph algorithm, transitive closure. We propose a disk-based transitive closure solution that operates on the bit-matrix. The solution is within the Python ecosystem while adhering to the principles of database systems. Our experimental study shows the superiority of our solutions over existing popular analysis systems, suggesting potential advancements in bridging high-performance computing and Python.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Houston

Year dc:date.issued

2024

Author dc:creator

• Zhou, Xiantian

Advisor dc:contributor.advisor

• Ordonez, Carlos

Committee members dc:contributor.committeemember

• Azencott, Robert
• Subhlok, Jaspal
• Huang, Stephen

• Graph analytics
• Big data
• Graph metrics
• Graph patterns

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10657/17736

OAI identifier oai:identifier

oai:uh-ir.tdl.org:10657/17736