Learning from Experience in Context of Climate Prediction in Pakistan using Data Mining Techniques Avidence from Computer Research n Pakistan

Abstract

The purpose of this paper is to analyze the climate changes in Pakistan, identify issues related to weather disasters and to revisit weather prediction approaches. The proposed approach is based on different algorithms and their comparisons with reference to past 5years (2010 - 2015) data on 12 attributes. A flow diagram is given that identifies steps included in the process. Results are obtained using WEKA 3.7.13 (latest version 2015). The KNN algorithm and memory-based reasoning algorithm shows the accuracy of predicting weather forecasts. The BPANN algorithm is used to analyze the data set along with KNN and memory-based reasoning algorithms. Decision tree shows the accuracy of predicting weather forecasts. The KNN is used with Bayesian approach in this research. Attributes used in this research shows significant relationship while many of those work as independent variables. Since, for weather prediction these attributes are very important, we used variant factors based on time and date. The KNN algorithm using Bayesian classifier provides accurate results compare with memory-based reasoning of Decision Tree and BPANN trainlm and trainbr.