Periodic Outlier Pattern Detection And Boost Prediction In Time-Series Data
Ankita Karale, Sandip M. Walunj · 2016
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.