Abstract:
According to figures from the Sri Lanka Tea Board, the amount of tea exported fluctuates
greatly depending on the time of the year. Analyzing and predicting tea export
based on the type of tea exported would be highly beneficial for various stakeholders in
the industry. Investigating the relationship between tea export and important factors
aids in identifying the elements that contribute to fluctuations in tea export volume.
Accordingly, this study focused on the collection of data specific to the future prediction
and analysis of tea export in Sri Lanka in terms of influencing factors and to analyze
them based on a thorough identification of existing research gaps. Monthly tea export
data over the past ten years, as well as prices and volumes of various tea types over
that time period were used in the study. These historical data were used to assess and
determine the significance of the correlation between the key factors and their variation
patterns in order to forecast tea export volume. This study used a variety of prediction
and forecasting methods with the Multilayer Perceptron, a type of feedforward Artificial
Neural Network emerged as one of the most effective methods for developing accurate
prediction models. The accuracy of the results was tested and evaluated using the
confusion matrix. The prediction model yielded an accuracy of 98% with a mean absolute
error of 0.02%, root mean squared error of 0.12%, 0.985 precision, and 0.984 for both
Recall and F-Measure. The study further demonstrated that the identified factors have
a satisfactory level of correlation in determining the tea export in Sri Lanka, with the
year, month, and tea type having the highest influence.
