Periodic Outlier Pattern Detection And Boost Prediction In Time-Series Data

Abstract

In todays era the data mining field has been studied allencompassing.
Outlier pattern detection as a
subdivision of data mining is a interesting problem and
has huge number of application. Outliers are nothing
but uncommon patterns that rarely occur. That’s why it
does not have proper support in the data. We can’t
consider the outlier patterns as noise though they appear
different with respect to all other patterns. Surprise
patterns may proposition toward variance in the
datasets. They can be transactions which are
fraudulent,customer behavior change, network
encroachment, recession which occur in the economy,
terrible weather conditions, etc. That’s why conclusion
may be drawn as periodicity detection of surprise
patterns may be more crucial in many series rather than
the regularity of regular patterns. The proposed system
will give us a solution for detection of outlier pattern.
This system will also make prediction about the future
coming outlier patterns.