Study and Analysis of Time Series of Weather Data of Classification and Clustering Techniques

Abstract

AbstractClimate on Earth is chaotic; hence, the meteorological department faces though challenge of forecasting weather to maximum accuracy. Weather forecasting is very important as it is the first step towards the preparation against the upcoming natural hazard and hence causing threat to life form. The objective of the paper is to study weather parameters using different clustering and classification techniques and select the best-suited model on the basis of least error and less time taking. Weather parameters such as max. and min. temperature, wind Speed, rainfall, relative humidity, evaporation, bright sun shine hours, average temperature, average humidity from 1 January 2017 till 30 September 2018 of Delhi region are studied. Navie Bayes provides better results as compared to others on the basis of statistical outcomes of Kappa Statistics, MAE, RMSE, RAE, RRSE. Clustering is the technique used for clustering the weather parameters into groups having similarity, and classification refers to the forecasting of weather parameters on the basis of the input data by training, validating, and testing the data set and further forecasting. K-means clustering, EM clustering, hierarchical clustered are the clustering methods used for study, and on the biases of the less time taken, it is concluded that K-means clustering is efficient clustering technique.KeywordsClassificationClusteringNavie BayesJ48LivSVMK-means clusteringEM clusteringHierarchical clusteredData mining