Investigation on factors affecting price of tea in Sri Lanka

Abstract

The thesis deals with an attempt to investigate on factors affecting price of tea in Sri Lanka. According to the Statistics & Information Technology Unit of the Sri Lanka Tea Board, the price of tea depends on many factors including crude oil price variation, US Dollar exchange rate variation, tea production, native exports, area harvested and climate seasons. The price of tea keeps varying with time and the changes in price are a worrying phenomenon for those involved directly and indirectly in the tea industry. Researches performed to analyze the correlation between the factors stated and the price of tea have produced varying and ambiguous results. According to Sri Lanka Tea Board, no appropriate tool is used in Sri Lankan tea industry to predict or forecast future price of tea based on environmental and economical factors. Problems related to tea price variations are to be overcome by focusing on investigating the correlation of the price of tea with the six important factors. Making use of historical data available and the knowledge discovery in databases model, the author concentrates on determining the factors that best predicts the price of tea and providing an explanation on the type and strength of the correlations. Multiple regression, a data mining technique that comes from a statistical framework not only investigates on the factors using an attribute selection method, but uses the selected factors to perform a test prediction on the price of tea. Performance evaluation measures used to test the accuracy of the attribute selection and the regression model showed a high correlation coefficient between the predicted price and the actual price with minimal error. The predictions performed using the selected factors were highly accurate according to the evaluation results. These findings including the selected factors and their impact on different categories of tea will be useful in predicting the price of tea in future.