Poster for: Approaches to systematic transaction data reuse: machine learning support of information discovery

Hahn, James F.

Permalink

https://hdl.handle.net/2142/103843 Copy

Description

Title

Poster for: Approaches to systematic transaction data reuse: machine learning support of information discovery

Author(s)

Hahn, James F.

Issue Date

2019-05-13

Keyword(s)

• machine learning
• network science
• FP-growth
• WEKA, Gephi
• personalization
• recommender systems

Abstract

The purpose of this poster is to detail approaches in systematic transactional data reuse using machine learning and network science visualization methods. Preliminary machine learning workflows were undertaken in WEKA and made use of the Fp-growth algorithm for seeding a recommender system for library user accounts. The resulting tables of consequent association rules are stored in a production relational database server accessed through the recommender app’s middleware. The association rule database is used at runtime for the machine learning based recommender system. As a result of the availability of databases with association rules, researchers further analyzed their properties using network analysis software were the edges of the graph are the consequent topics and the nodes are the antecedent topics. Network science can be particularly valuable in assisting the understanding of machine learning outputs by visualizing and analyzing graphs the distribution of the topic network. Several key network science metrics, including the average degree, diameter, average clustering coefficient, and total triangles were computed by way of Gephi network analytic tools.

Publisher

F1000Research

Type of Resource

• text
• image

Language

en

Permalink

http://hdl.handle.net/2142/103843

DOI

https://doi.org/10.7490/f1000research.1116746.1

Sponsor(s)/Grant Number(s)

University of Illinois Library, Research and Publications Committee Award 81034

Copyright and License Information

• © Jim Hahn 2019
• This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0)

Owning Collections