1. TVCG 2012 Bilal Alsallakh, Wolfgang Aigner, Silvia Miksch, and M. Eduard Grïoller

2.  Introduction  Related work ◦ Contingency Wheel  Contingency Wheel++  User Study  Conclusion

3.  Contingency table ◦ a common way to summarize categorical data as a first step of analysis ◦ an nxm matrix that records the frequency of observations ƒ ij for each combination of categories of two categorical variables  Data ◦ about 1million user ratings on 3706 movies  a 3706x21 table which counts for each movie, how many times it was rated from users of each occupation  a 604017 table which counts for each user, how many times he/she rated movies from each genre

5.  the column categories are visualized as sectors of a ring chart  the table cells are depicted as dots in these sectors  the dot for cell (i,j ) is placed in sector i at a radial distance from the ring’s inner circle proportional to the strength of association r ij between row i and column j  to reduce the large number of dots ◦ by filtering out cells (i,j ) with r ij ≦T r (where T r is the association threshold) ◦ by filtering out entire rows with ƒ i + <T s (where T s is the support threshold)

7.  Data mapping ◦ to visualize association values r ij  Visual mapping ◦ it difficult to accurately interpret the meaning of these dots at the beginning  Interaction ◦ dots closer to the center were often too small and overlapping

9. where cte is a constant computed from the table to ensure -1≦ r i j ≦1

11. Distributions of (a) a numerical attribute (release date) or, (b) a categorical attribute (genre) of the movies in the histograms. (c) The global distributions of release date and genre among all movies.

16. Details about selected genres.

17.  a novel visual analytics methods ◦ visualize and discover patterns in large categorical data  improve Contingency Wheel  offer a multi-level overview+detail interface