Business Intelligence Integration Joel Da Costa, Takudzwa Mabande, Richard Migwalla Antoine Bagula, Joseph Balikuddembe.

Business Intelligence IntegrationJoel Da Costa, Takudzwa Mabande, Richard MigwallaAntoine Bagula, Joseph Balikuddembe

Business Intelligence

•What•How•Why

Current BI Trends

•Predictive Analysis•Real-Time Monitoring•In-Memory Processing•Software as a Service

Problem Statement

•Previously ‘one size fits all’•Which are actually the most effective ?

▫Bayesian Belief Networks (GA)▫Neural Networks (GA)▫Artificial Immune Systems

Cases

•Profiling Customers•Predictive Sales Forecasting

Aim

•See variance of results on same data•Define strengths and weaknesses in BI

technologies

ApproachBrief Look into the rationale behind our proposed solution

Overview

(Yet to add Diagram here…)

Input

•Previous Works▫S. Mahfoud and G. Mani▫P.-C. Chang

•Sanlam Specification▫Sales▫Income

Interface

•Simplified Interface▫Graphical Display ▫Relevant information▫Technical Data Hiding

System Approach 1:

Bayesian Belief Networks• Joel De Costa

(Diagram here)

System Approach 2:

Neural Networks (NN)• Takudzwa Mabande

System Approach 3:

Artificial Immune Systems (AIS)•Richard Migwalla•Overview

▫Abstraction of Human immune System(Diagram here)

Output:

•Sanlam Specification▫Predicted sales▫Customer Profile

Likely Purchase based on current income

Division Of Work

Bayesian Networks

Joel

Neural Networks Takudzwa

Artificial Immune SystemRichard

Connecting To

Database Joel

Customer DB

Interface

Richard

Sales DB Interface Takudzw

a

Sales & Customer

Visualisation

Takudzwa

GUIRichard

Timeline

RisksRisk Matrix Evaluation Avoidance Mitigation

1.

Loss of a project team member.

D. Serious/ Low Probability

Pressure to stay on the project as failure to do so means not graduating.

Have sufficiently independent deliverable modules for each team member.

2.Delay in Delivery of test data.

C. Disastrous/ Low Probability

Pressure Sanlam to provide data as soon as possible.

Create random test data or use alternative available data.

3. Scope creep (Plan too many tasks, Cannot complete tasks in time)

E. Marginal/ Low Probability

Project planned in detail with supervisor and department approval.

Start with fundamental features first and leave other things to the end.

4.Data loss due to hardware failure, (External Factor)

C. Serious/ Medium Probability

Frequent backups of all progress on different machines or storage devices.

Roll back to last backup.

5.

Missing project deadlinesC. Serious/ Medium Probability

Constant reference to the project timeline and clear communication between project members

Review and reassess deadlines; readjusting where necessary- as cost-effectively as possible.

6. Misunderstanding User requirements. (Resultant of miscommunication/ ambiguity in user-team interaction)

D. Serious/ Low Probability

Constant communication with Sanlam and providing them with project plan and design in order to detect flaws.

Iterations through development so that inconsistencies can be detected early.

Resources

•Lab PC’s•Access to Sanlam Database•Java Development Enviroment•Project team

DB

Anticipated Outcomes

We will create a package that will: •Read in data from the Sanlam database. •Use different machine learning

techniques to profile customers and forecast sales.

•Compare the accuracy of the different techniques using actual data.

•Identify the best technique for use in each particular scenario.

Key Success Factors

•Identifying the best technique for Customer Profiling

•Identifying the best technique for Sales Forecasting

•All techniques performing approximately the same amount of work (i.e. same data, about the same time, relatively the same complexity)