Mulyaankan: Discrepancy detection using data mining
Mulyaankan is designed to aid discrepancy detection in export/import valuation. Importers, often undervalue their goods, so as to evade or, to lessen the burden of import duty. Humans are able to detect discrepancies in the declared price, by comparing various properties of one import vis-a-vis others in the same category.
Human intuition works by trying to form various patterns in the given data and comparing data against the rules specified by the patterns. Any item with a declared price, out of the ordinary, could be flagged as suspicious and forwarded for further investigation. Mulyaankan aims to mimic this behaviour over the large dataset that the Customs department handles, thus providing the user with the freedom to concentrate on the suspicious items, rather than focus on extracting the patterns from the mind-boggling amount of data generated at various customs houses.
In its final avatar, it aims to be a risk profiler, with the ability to take in data from the database and then, based on patterns previously found, identify importers whose declarations are likely to be inaccurate.
Currently, Mulyaankan acts as a repository for all the import data in the country and performs statistical analysis on it. Periodically, data is fed into the system and it generates a report for all the Customs Stations (or field formations in Customs' lingo) of the current prices for all the commodities imported in the country, along with their price range and international prices. It also provides an indication of the products/imports, that it considers to be 'Odd', thus drastically cutting down on the number of cases to follow up manually (the decrease is sometimes of the order of 100!). This discrepancy detection is rated by the Customs Department as the single most valuable feature in Mulyaankan. Given the humungous number of products being imported, it is infeasible to analyse all of them, hence the Department of Valuation prepares a list of commodities it is interested in and creates a profile for each one of them, at a fine level of granularity. Mulyaankan takes these profiles and performs analysis on the incoming data only for the items mentioned in the profile.
There are elaborate querying mechanisms, meant for a whole range of users from the beginner to the technically proficient one. Combined with flexible filtering abilities, this enables the Mulyaankan User to go from a bird's eye view to worm's eye view, in a jiffy! A nifty graphing capability adds to the range of views of the data that is made available to the user. All in all, Mulyaankan aims to reduce the flood of data into actionable information for the Directorate of Valuation, inside the Customs Department.
Technical tidbit: Mulyaankan has been developed using C++ and MFC. The data is kept in an Oracle database.
Mulyaankan team can be reached at : mulyaankan [at] cdacmumbai [dot] in
Staff members currently associated with project Mulyaankan : Dr. M Sasikumar, Mrs. Veena Tyagi, Ankit Dangi
