Data Mining Approach for Studying the behaviour of Weather Variations in Sri Lankan Context

Abstract

Weather forecasting is the task of determining the future conditions of the atmosphere for a given area. It is a one of the most challenging issues around the world for more than decades. Accurate weather prediction is more important because it has a direct impact on the social and the economic factors. For an example agricultural and industrial sectors are heavily depending on the weather predictions. Since Sri Lanka is a tropical country, it consists with mainly two seasons such as dry and wet season. These seasonal differences are highly impact on the economy of the country because agricultural products and rice production are act as the major role in Sri Lankan economy. Therefore, a reliable weather prediction is necessary to determine the best time to start planting and gaining maximum harvest. Further the meteorological elements such as rainfall, temperature, humidity and windspeed are immensely affect many aspects of human livelihood. They provide analytical support for the issues related to urban computing such as electric power generation planning, traffic flow prediction, air quality analysis and so on. Therefore, in recent weather prediction has become a more important research area. Hence the researches are more concern with developing a reliable and accurate weather prediction model. So, the main goal of this research is to estimate the weather variations by utilizing the predictive analysis. During this analysis, various data mining and machine learning algorithms are used to develop a better weather prediction model. This research mainly introduces the Artificial Neural Network, Time series analysis, Regression analysis and Decision Tree approaches as the main data mining and machine learning techniques for predicating the weather conditions. Throughout this research it mainly concerns about the rainfall, temperature, windspeed, relative humidity and atmospheric pressure as the weather parameters for developing the predictive model. The field of machine learning has received much interest from scientific community. Hence machine learning techniques like artificial neural networks are a good candidate for the prediction of weather conditions with large data sets. Also, weather forecasting with time series analysis has become an important mechanism in numerous meteorological applications as well as other environmental areas for determining the phenomena like temperature, relative humidity, pressure, rainfall etc. Therefore, the major concern of this research is to develop a weather prediction model using artificial neural networks and time series analysis while applying the other data mining and machine learning approaches such as and regression analysis and decision trees to achieve a reliable forecasting model.