Statistical network analysis concepts

This PhD follows statistician John Tukey s concept of Exploratory Data Analysis [12] the primary purpose of visualizing and exploring is to raise questions and gather insights about a large quantity of data. Unlike most statistical work, which evaluates a priori questions according to a model, exploration by information visualization has the potential to start analysis without assumptions, or open new perspectives on a previously-analyzed dataset. For these purposes, overviews of the whole network are crucial. 

The paucity literature reported on applications of ANNs in marine and offshore safety engineering reflects that the concept of ANNs is still an extremely raw technique to this area. Published research literature providing a step by step explanation of input data identification through network architecture design and output analysis is somewhat sparse. Buxton et al. (1997) applied the techniques of ANNs to statistics of losses of bulk carriers due to fire to determine whether it is of potential value as a predictor of overall risk. More recently, some initial findings based on a feasibility study of using ANN techniques in offshore and maritime safety-based decision support system has been reported (Sii et al. (2000), Wang et al. (2001)). 

---