Predicting The Market Status Of Coffee, Pea Beans And Sesame: The Case Of Ethiopia Commodity Exchange
No Thumbnail Available
Date
2015-10-02
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Many commodity exchange organizations; including the Ethiopia Commodity Exchange (ECX) have been trying to find a way to predict future market status (in terms of price and sales volume) for commodities that are being traded within the exchange’s trading platform. Data mining helps to find predictive information from large databases and data warehouses. Companies use predictive modeling tools for strategic decision-makings and by analyzing the company’s historical information we can anticipate the changes in the future. Thus, in this study, by analyzing the historical ECX market data (EMD) obtained from ECX; we have discussed on different data mining methods which are helpful in building a predictive data mining model.
The Hybrid Knowledge Discovery Process for Data Mining is followed to build the predictive model that analyzes and predicts the price and sales volume. This methodology was developed, by adopting the CRoss-Industry Standard Processes for Data Mining (CRISP-DM) model to the needs of academic research community. Our work is based on finding suitable data sets as well as best predictive model that helps in achieving high accuracy and generality for Weighted Average Price (WAP) in Ethiopian Birr and Sales Volume in Tons predictions for selected agricultural commodities (Coffee, Pea Beans and Sesame) traded at ECX. For solving this problem, different data mining classification techniques (Eight WEKA classifiers: Gaussian Processes, Linear Regression, Multilayer Percepton, SMOreg, Decision Table, M5Rules, M5P and REPTree) were evaluated on different data sets.
The experimental results obtained from this study shows; the Decision Table Classifier has the highest optimal accuracy score with maximum optimal Correlation Coefficient Percentage (CCP) of 97.8%, minimum optimal Root Mean Square Percentage Error (RMSPE) of 3.9% and an optimal Time of 0.02 seconds to build the Price Predictive model. While, the M5Rules Classifier has the highest optimal accuracy score with maximum optimal CCP (80.5%), minimum optimal RMSPE (9.4%) and an optimal Time (0.11 seconds) to build the Sales Volume Predictive model.
Finally, by extending WEKA software source code, an application (predictive-model-prototype) which is termed as “ECX Price and Sales Volume Predictive System” with a user-friendly GUI is developed and deployed for the usage of domain experts (end users). Therefore, the results obtained from this research indicate that data mining classification models are very useful in predicting price and sales volume of commodities traded in commodity exchange organizations like ECX for the effective and efficient utilization of massive amount of market data to support experts and stockholders (traders) in making strategic planning as well as proactive and knowledge-driven decisions.
Description
Keywords
Market Status Of Coffee, Pea Beans And Sesame: The Case Of Ethiopia Commodity Exchange