Development and Implementation of the Predictive System of Zambian Kwacha Rate using Data Mining

This research aims to analyze the daily data of the Zambian kwacha exchange rate against the US dollar from January 1st, 2011, utilizing various data mining models to identify the most effective approach for price predictions. While traditional methods such as purchase power parity, relative economic strength, and econometric models are still employed in forecasting exchange rates, there is a need to explore machine learning concepts to improve prediction accuracy. The research methodology follows the Cross Industry Standard Process for Data Mining (CRISP-DM), consisting of six phases: business understanding, data understanding, data preparation, modeling, evaluation, and deployment. Various data sources were collected, including datasets from the Bank of Zambia website, and data exploration techniques were employed to uncover patterns and insights within the data. Different data mining models were applied, including decision trees, linear regression, random forest, and ARIMA models. Decision trees provide a non-parametric supervised learning method, while linear regression establishes a relationship between variables. Random forest is an ensemble technique that combines multiple decision trees, and ARIMA models incorporate information from leading indicators and exogenous variables for enhanced forecasting accuracy. By utilizing CRISP-DM and employing data mining models, this research aims to contribute to the understanding of factors impacting the Zambian Kwacha exchange rate and identify the most effective approach for currency prediction. The findings can potentially assist in improving exchange rate forecasting and inform decision-making in the financial and economic sectors.