By this point in the semester, everyone’s looking forward to the mid-semester break, and I’ve relocated to the lecture series in the Scientia building during the warmer hour of 1-2PM. Patrick is lecturing on Business Intelligence (BI), which I shall retrospectively define as the knowledge produced by a Decision Support System (DSS), which is an information system designed to assist decision-making. We love our definitions in INFS1602! We also love our YouTube videos, and hence the lecture slides are quickly Alt+Tabbed to a YouTube video about a credit card company that uses BI.

Now, let’s take it easy by drifting into shallow waters to talk about decision theory. We begin with the structure continuum (my HSC IPT teacher coined this term – I like it, and old habits die hard). Problems can be categorised as structured or unstructured depending on whether or not they can be solved through an algorithm (sequence of steps) given parameters. If they can’t, they’re unstructured, usually because of uncertainty. But most problems are semi-structured, of course. Decision theory continues with Herbert Simon’s decision-making process:

1. Intelligence – gather the relevant facts and data
3. Choice – decide which path to follow

Patrick even touches on the concept of satisficing at point 3 – “Could you waste 2 hours looking for the best lunch, or would you just want a lunch?”, he asks in his lovely Irish accent. The definition of a model as ‘a representation of reality’ is given with minimal fuss.

The lecture now drifts to deeper waters with the technical side of things. The anatomy of a DSS is given as {data management, model management, dialog/UI}. We know what a model is – he just gave us the definition – but to clarify exactly what data management is, we meet the data warehouse, and its younger sister with braces, the data mart. It is here that Patrick shows us a diagram so helpful that it was repeated in at least two tutorials. As such, I have replicated it with an example from my own life, a Decision Support System about my physical fitness based on body metrics and running stats:

OLAP and Data Mining are mentioned in that diagram. This is where the lecture heads to next. Firstly, OLAP, represented by a ‘cube’, refers to querying your data warehouse in multiple dimensions. SQL and INFS1603 come to mind. Data Mining, on the other hand, is more high-tech, artificial-intelligence, fancy-pantsy stuff. It’s about finding hidden relationships in your data. Patrick gives the example that Walmart allegedly used data mining to discover that nappies and beer would be a great combo because dads buy both in the afternoon.

The tutorial rolls along and Tony, our tutor, introduces us to Pivot Tables after discovering that nobody in the class had heard about them. Pivot Tables are essentially an OLAP tool built into Microsoft Excel (and other spreadsheet software). And they are awesome:

(Months on the row labels, years on the column labels. For example, it is currently the 6th month of 2012 and I have run 36 kilometres this month so far.)

The prescribed case study for this week is about the Netflix Prize, which Wikipedia defines as “an open competition for the best collaborative filtering algorithm to predict user ratings for films, based on previous ratings” (this is vastly superior to the textbook’s overview). Interesting points are made during the tutorial about open innovation, peer production, and privacy issues.