Detection of Rare Patterns in Climate Change Using Data Mining Techniques

Mr. Parag N. Kolhe, Mr. Rahul M.Ugale and Miss Shraddha V Shingne

Today the volume of data has been enormously increasing as a result of advances in data generation, collection and storage technologies. The effect of climate prediction on society, business, agriculture and almost all aspects of human life, force the scientist to give proper attention to the matter. Weather is a continuous, data-intensive, multidimensional, dynamic process that makes weather forecasting a formidable challenge. The prediction of rare events is a pressing scientific problem. We primarily concentrate to identify weather patterns in the long term while consistent with global climate change on weather patterns, identify rare/outlying patterns that coincide with rare events data mining.

This paper propose an adaptive clustering pattern detection method for the detection of rare patterns in climate change using data mining techniques which uses k-means algorithm where an open number of states as clusters to accommodate the dynamic temporarily of data. By adding adaptive clustering property as a global restriction, the granular size of the clusters is determined for optimal performance. The global modeling result is presented which provides a base of data mining tasks.

The metrological variables longitude, latitude, mer wind, zone wind, humidity, air temperature and sea surface temperature are analyzed to detect climate change patterns in this study. The result depicts different patterns of climate in the form of histogram based on the records of metrological variables of NOOA. Different distance measures are applied between the centers of the clusters formed for testing the sensitivity of the method. The robustness of our method is demonstrated by the results. Our method of detecting rare patterns in climate change will be very useful for weather and metrological research focusing on the trends in weather and the consequent changes. Adaptive clustering method uses an open number of states as clusters to accommodate the dynamic temporarily of data.

By adding adaptive clustering property as a global restriction, the granular size of the clusters is determined for optimal performance. The global modeling result is presented which provides a base of data mining tasks. This adaptive climate change pattern detection algorithm will be proven to be of potential use for climatic and meteorological research as well as research focusing on temporal trends in weather and the consequent changes.

©2015 International Journal of Computer Science and Applications 

Published by Research Publications, India