Development of intelligent hybrid learning system using clustering and knowledge-based neural networks for economic forecasting : First phase

• Main Authors: Che Mat @ Mohd Shukor, Zamzarina, Md Sap, Mohd Noor
• Format: Conference or Workshop Item
• Language: English
• Published: 2004
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://repo.uum.edu.my/13907/1/KM184.pdf; http://repo.uum.edu.my/13907/; http://www.kmice.cms.net.my
• Institution: Universiti Utara Malaysia

The economic forecasting environment is currently undergoing drastic changes and has a complex and challenging task.Practically, people design a database application or use a statistical package to conduct the analysis on the data.Former approach can be done on the online data, but it must be developed after stating the goal of analysis, which means it only possible for a limited and specific purpose.Whereas the statistical approach must be done for the offline data, however it can lead to the missing pattern and undiscovered knowledge from the available data (Shan, C., 1998).For the effort to extract implicit, previously unknown, hidden and potentially useful information from raw data in an automatic fashion, leads us to the usage of data mining technique that receives big attention from the researchers recently.This paper proposed the issues of joint clustering and knowledge-based neural networks techniques as the application for point forecast decision making.Future prediction (e.g., political condition, corporation factors, macro economy factors, and psychological factors of investors) perform an important rule in Stock Exchange, so in our prediction model we will be able to predict results more precisely. We proposed KMeans clustering algorithm that is based on multidimensional scaling, joined with neural knowledge based technique algorithm for supporting the learning module to generate interesting clusters that will generate interesting rules for extracting knowledge from stock exchange databases efficiently and accurately.