Building a basic AI algorithm - Microsoft Excel Tutorial

- [Instructor] One common application for expert systems and AI in general is in the financial field. You might be familiar with tools called Robo Advisors. I'm in the 03_03_Begin Excel file. Now, this is an example of a basic Robo Advisor that we built through Excel. All this really does is it asks the investor a series of questions and then makes a recommendation based on the responses of that investor. So how does this work? Well, it revolves around a basic algorithm running in the background. So if we look at the portfolio calculator sheet, you'll see that we take the customer's responses and we translate them into numbers. Once we have those numbers, those responses, we're going to turn around and come up with an allocation. How do we do that? Well, it's built on basic programming language in the background combined with a set of facts or data that we've already pre-input. Essentially, if we think about this in the context of the expert system, this if statement, or vlookup statement in this particular case, is our logic; it's our inference engine. The tools that it's relying on behind the scenes, what we're calling parameters for the system over here in red and green, that's our knowledge base. Now notice, when we get done, for this particular customer, the system is going to recommend 51.3% allocation of assets to stocks. Well, how does it come up with that? It's based on the choices the investor put in, but more importantly, the data that we have in the background. If we adjust our knowledge base and say now that, as an example, we should have a 25% allocation for age number one, what happens? And we're going to go through and adjust to a 10% for age two, and let's say a negative 10% for age three, and a negative 25% for age four. Well now look what happens. Our stock allocation jumps up to 56.3% and our bond allocation falls correspondingly. We could also go through and change our risk adjustments, so that maybe they reflect some other state that we've decided is more optimal. And I'm just picking numbers here, but you get the point. Any of these adjustments that we make to the facts behind the scene, flow through the inference engine and change the output that the investor sees. And all of that's reflected in a changed allocation or recommendation to the investor. Now there's nothing especially difficult here, but the key point is if we're going to use AI effectively, we've got to have these parameters set correctly. Where do those parameters come from? Well, it's largely going to be based on historical data, but remember there's a mass of historical data out there. Humans have to go through and cull that data and figure out what data to show to the computer. So as exciting as AI is, ultimately at this stage it's still very much dependent on human intuition and decision making. Assuming that we get those facts right behind the scenes, that's going to spit out our allocation, and then we'll translate that into a set of expected returns for the investor and a set of recommendations for how they should allocate their money. Now this is simply a basic expert system, drawn from very little data, as you'll see here, but nevertheless it can be very effective for helping with a very common problem: how should we invest, given a pool of money and given certain needs that the investor expresses.