Modified Projected Space Clustering Model on Weather Data to Predict Climate of Next Season

Objectives: Four seasons of Indian weather are interdependent. Prediction of seasonal weather supports many fields to work successfully. The objective of proposed model to work on Weather Pattern identification in Initial phase of prediction which has to include unequal weight of attributes. Methods: The projected space clustering model is suitable to handle the non-sequence patterns of data set. The existing projected space clustering eliminates the least weighted attribute. The framework suggested in this paper incorporate modified projected space cluster which work on complete set of attributes to form pattern wise clusters which is dynamic in number for each season. Next part of framework is seasonal weather prediction using ANN, works on dynamic set of clusters. Findings: Dynamic nature of clusters formed in modified projected space clustering completely eliminates the error rate arise because of fixed number of cluster. The extreme events patterns formed as a separate clusters are not eliminated as outline. The result of these clusters gives the study report of each season, like the changes of climate pattern, the frequency of extreme event and weather prediction of next season. Application/ Improvement: The modified projected space clustering work well on unequal complete set of attributes to form a cluster of different pattern. For each duration numbers of clusters are dynamic based on the pattern variation in climate data.