10-04-2017, 07:55 PM
C-TREND : A new technique for identifying trends in transactional data.
Identifying the temporal relations in data is a major problem in many business and academic domains. Time series analysis, keogh and kasetty etc are some specialised techniques used. The advantages of finding the temporal relationship can be forecasting and making decisions in future investment patterns and business strategies. For example, forecasting of technology analysis may require the analysis of more than one parameters at once.
Approach
The C-Trend separates data into user defined partitions on the basis of the time period. And from this the clusters of dominant transaction types taking place in each partition is identified. Then the clusters from a particular time period is compared with the cluster in the nearby time period. Cross period similarities are identified and from long time data , the trends are identified. he two main processesw involved in the technique are:
-offline preprocessing of the data
-online interactive analysis and visualisation of the trends.
For further details, visit:
http://citeseerx.ist.psu.edu/viewdoc/dow...1&type=pdf