Design team member: Jeffrey Leung
Supervisor: Mohamed Kamel
Background
Data warehousing has become very popular among organizations seeking to utilize information technology to gain a competitive advantage. Moreover, many vendors, having noticed this trend, have begun to manufacture various kinds of hardware, software, and tools to help data warehouses function more effectively. Despite the increasing attention to data warehousing, there is a need for better data mining algorithms to discover certain patterns that would help the organization to make effective strategic decisions. Unfortunately, gathering information for a warehouse usually involves focusing on a specific data domain. For example, the business sectors have data warehouses that contain the data of the factors affecting sales growth. In order to discover patterns that provide a broader view of an organization, one must include the data collected from multiple domains such as customer and employee satisfaction and the type of information exchanged between each party involving the organization. Therefore, one feasible solution to collect data from multiple domains is to refine the current Customer Relationship Management Model (CRM) to include data mining systems to build a broader perspective in the knowledge base.
Project description
The objective of this project is to develop a CRM application with data mining capabilities that allow the user to extract data from multiple domains to discover business knowledge. The project uses the business problems presented by Spectra Aluminum Products, a manufacturer of aluminum dies, as a case study to design the software solution. The software solution will be built using the latest software technologies such as DHTML, MTS, ADO and COM+, which realize the software architecture deployed by DieShop Organizer. The software architecture will follow the conventional three-tier architecture that allows the users of the system to access the features of the software from a web-browser and a desktop application. Attention to details in the design is important to DieShop Organizer which could be released as a commercial product.
The data mining system of DieShop Organizer will employ the web document clustering algorithms for categorizing the data and identifying certain patterns. Part of the data mining system will interface with the functionalities of OLAP in order to provide additional data mining features not supported by the web document clustering algorithms.
Design methodology
The design methodology of DieShop Organizer will followed the iterative software development process defined by the Rational Unified Process. The general steps taken by this approach is summarized below.
Research: The team performs a feasibility analysis on different types of software technologies available. This include COM+ and C# from Microsoft and Corba and JavaBeans from Sun Microsystems. Data mining clustering algorithms, such as the K-Means and Suffix Tree Clustering algorithm will be evaluated. A software requirement specification (SRS) is created to finalize the functionalities of DieShop Organizer.
Design / Implementation: In this stage, the team uses the Unified Modeling language to design the presentation, business and data objects. Rational Rose will be used as the designer tool to translate UML designs to code and reverse existing designs for design evaluation purposes.
Functional validation: This stage is used to test the software using the appropriate testing methods.
Re-define project scope: The team will require to re-define the scope of the project for the next iteration in the Rational Unified Process by evaluating the outcome of design, implementation and functional validation.