Uncertainty and Climate Change and its effect on Generalization and Prediction abilities by creating Diverse Classifiers and Feature Section Methods using Information Fusion

The model forecast suggests a deterministic approach. Forecasting was traditionally done by a single model deterministic prediction, recent years has witnessed drastic changes. Today, with Information Fusion (Ensemble) technique it is possible to improve the generalization ability of classifiers with high levels of reliability. Through Information Fusion it is easily possible to combine diverse & independent outcomes for decision-making. This approach adopts the idea of combining the results of multiple methods (two-way interactions between them) using appropriate model on the testset. Although uncertainties are often very significant, for the purpose of single prediction, especially at the initial stage, one dose not consider uncertainties in the model, the initial conditions, or the very nature of the climate (environment or atmosphere) itself using single model. If we make small changes in the initial parameter setting, it will result in change in predictive accuracy of the model. Similarly, uncertainty in model physics can result in large forecast differences and errors. So, instead of running one prediction, run a collection/package/bundle (ensemble) of predictions, each one kick starting from a different initial state or with different conditions and sequentially executing the next. The variations resulting due to execution of different prediction package/model could be then used (independently combining or aggregating) to estimate the uncertainty of the prediction, giving us better accuracy and reliability. In this paper the authors propose to use Information fusion technique that will provide insight of probable key parameters that is necessary to purposefully evaluate the successes of new generation of products and services, improving forecasting. Ensembles can be creatively applied to provide insight against the new generation products yielding higher probabilities of success. Ensemble will yield critical features of the products and also provide insight to forecasting ultimately improving the predicative skills & capabilities. This is accomplished by creative selection of multiple predicators and combining the same to crack down the complexity. Diversity can be achieved from different algorithms, or algorithm parameters.