Deep Data Mining with network relationships

Abstract

This thesis formulates a methodology that integrates Social Network Analysis (SNA) methods with traditional Data Mining (DM) methods. Traditional DM considers all the data as the independent entities and creates predictive models with data independence as an assumption. With Big Data evolving there is a need of much advanced data analysis techniques as the classical DM algorithms are ripe to change. This research demonstrates introducing characteristics of network relationship improves the accuracy of predictive models. This research add its value to the Deep Data Mining space. Traditional data mining models have been excellent to look at a single view of data and predict the next best outcomes. But with data science enhancement and the focus on data driven activities it is very important to understand the holistic view of data. Network relations help to understand the holistic view of data, whether it is customer centric or product centric or service centric. The research establishes a methodology that can be operationalised in any business domain such as scientific explorations of genes, better understanding of consumers and products to support demands of business culture. The methodology is demonstrated with the help of two case studies based on University publications data. To summarise, this research reflects on the capability to model and understand networks are fundamental to next generation advanced analytics. In the current competitive enterprise world there is a need of holistic view of data rather than a silo view. Understanding relationships between data entities could provide a holistic view on data. Mining on a holistic dataset has its potential to extract actionable patterns that can be beneficial for business, users, and consumers. The research presented in this thesis overcomes the assumption of DM that all records are independent by bringing a relationship dependence. The research methodology have reflected upon how it could solve business problems. In fact, organisations that require customer relationship management, product customisation, and superior customer experience may not be optimising these critical business capabilities unless they understand the networks that relate to these processes. An understanding of networks is at the heart of extracting insights from social media, delivering real-time marketing to individuals, and offering a superior customer experience. However, to model networks we need to understand characteristics of networks.

Date of Award	2016
Original language	English