This project was done as a part of Data Visual and Analytics course - CSE6242 at Georgia Institute of Technology.
A fake reviewer detection system using Belief propagation
Dependencies: numpy sumproduct
features.py -> function kde(): #input : grouped_df -> reviewer wise grouped data :
#output: grouped_pr -> product wise grouped data including burst periods in ordinal form:
product_id, ratings, date, bursts
5555991584, [5.0, 5.0, 4.0, 5.0, 5.0, 5.0], [[730243], [730261], [730267], [730411]], [[[730243], [735079]]]
To convert dates from ordinal to date format, use fromordinal()