Introduction: Abstract: How to determine an appropriate view of data analysis processes and data presentations is an extremely important issue when people want to understand trivial data or common-sense concepts as results of intensive data analyses. This issue involves a precise granularity identification as well as a flexible shifting between the different layers of views, fitting various settings and satisfying different users demands. The data analysis results of a transient change analysis of traffic flows of all crosses in a large city within a minute at a morning peak time might be useless to a car driver trapped in a traffic congestion when going to work. Likewise, it does not help very much to this driver either if a general city congestion index for a day is provided to her. However, this situation happens from time to time when researchers apply data analysis methods such as Predictive Analytics (PA) or Artificial Neuro-network (ANN) but do not realize an appropriate granularity of semantic views should be considered and figured out, resulting in difficulty for the users to understand and use the big data analysis outcomes. In this talk the author intends to point out what various data analysis methods are suitable for which views in which an appropriate granularity is defined. The author like to emphasize that, for the last few years, lack of deep analysis of semantic analysis and conceptual modelling in the research field of big data analysis for the areas such as e-commerce, social network systems, public traffic systems, and general healthcare, where huge amount of data accumulated in volume along the time is a major cause that results in this granularity problem.​