A synergistic paradigm for intelligent multivariate data classification

The main objective of this study is to develop an intelligent data classification model that integrates techniques from Statistical, Neural Networks, Machine Learning and Knowledge Based Expert Systems approaches. The goal of such synergy is to overcome the limitations of the various approaches and improve data classification results. The model suggested stems from an information systems view of data classification and is designed as a foundation for an intelligent decision support system that can assist decision makers especially in a data intensive management environment. The model is developed and tested through several phases. First a conceptual framework representing the synergistic approach is built. Second a methodology for developing the system is designed. Third the model is developed. The model uses the results of a logistic regression classifier to feed a Neural Network (Backpropagation) and a Machine Learning Algorithm (ID3). Results from these two classifiers are then integrated and interpreted using a Knowledge Based System. Finally the model is evaluated comparing its efficiency to that of other classifiers. The efficiency is measured by classification accuracy (the percentage of correct classifications on new cases) and reliability (the variance of classification results on test samples). The model is evaluated using samples from actual customer's data collected by a national service company. Results have shown that this approach can significantly improve the accuracy and reliability of classification while also providing interpretation to the results. To extend its results to other domains, further testing using different data is recommended. This study contributes to the ongoing research in improving data classification and provides a structured methodology for implementing the suggested model.