Predictive Analytics

By Stephen Downes
Oct 22, 2021
Transcript of Predictive Analytics

Autotranscribed by Google, unedited.

Hi. I'm Stephen Downes. Welcome back to another session of ethics analytics and the duty of care, module two. And in this video, we're going to look at the subject of predictive analytics. We've been looking at different types of learning analytics. So far through this module, we've already covered descriptive analytics and diagnostic analytics and this is the next one of six different types of analytics that we're looking at.

Predictive analytics essentially involves do a two-stage process. And the purpose of them is to answer the question of what will happen in the future based on an identification of trends and patterns in existing data. So the first phase is to identify the patterns in the trends in the existing data and that's a lot of what we are doing with descriptive and pre and diagnostic analytics.

The second stage which is new is that we take the predictive model that we created in the first stage and we add new data to it and the outcome of that application is a prediction of some sort. As you may imagine predictive analytics have wide, a wide range of uses in online learning, and learning technology will be sampling.

A number of them here. In this presentation one such is resource planning, resource planning is important. Of course to educational institutions. They worry about everything from the number of staff to have the number of classrooms to have available the number of books to put in the bookstore. And in the example discussed here predicting website traffic on their website, to make sure that there are no server issues or even doing things to ensure campus health, such as working with Twitter data to predict outbreaks.

A major application of predictive analytics, is learning design. And in this example here we have a case where some researchers linked together. A hundred and fifty one different modules taught at the open university and used them. Use the learning design in order to predict whether the learning design had any impact on their behavior and ultimately on their success in the course.

And as you might expect from this study, there was a relation. So, it now becomes a mechanism for creating a learning design and then being able to predict whether that learning design will actually be useful in the context of an online. Course, obviously many other types of analytics are used in learning design as well.

This specific application is used for testing of proposed learning designs. Another example of testing is the use of predictive analytics in user testing. It's a similar sort of approach as the one for learning design. The idea here is that we want to be able to predict whether a person will use a website, a certain way, for example, in this case tongue and his colleagues predicted, whether a person would start or stop watching videos.

And so you do this user testing and you come up with these predictions for different types of videos, different types of materials or different types of website design, another type of predictive analytics and one of the most widely talked about types of predictive analytics in the field is that used to identify students at risk of failing.

We can imagine this being done based on very simple tests. For example, if a student never attended any classes, then the odds are greater that they will fail. But what about many other criteria? That might have an impact on a student's potential passing or feeling in the class. Here, we have an example where we look at everything from household income, to parents marital status to medical conditions to great point, average to neighborhood demographics.

All of these factors taken together can help an institution predict, whether a student is at risk of failing or not. This in turn helps institutions with another application academic. Advising there's an opportunity for advisors to incorporate elements of AI to their toolkits. We read allowing them to free up, time to form personal relationships with their students.

The idea here is that the advisor spends less time, trying to figure out what approaches will be most successful. What factors are involved in the student being likely to pass or fail succeed or not? And it allows them to use that analytics in the background to inform and help their personal interactions with the individual student.

There's a field in fact, out there called precision education, and if you do a search on this, you'll find a number of resources yang and ogata, right? That the goal of precision education is to identify at risk students as early as possible and provide timely intervention, based on teaching and learning experiences.

We can see why we've grouped all of these into the same category, they're all doing the same sort of thing, looking for or identifying patterns, in a student's background behavior circumstances, environment, etc. In order to arrive at some sort of prediction as to what they will do, whether they will pass or fail, whether they will use our website a certain way, whether they will watch a video to completion etc, the same sort of approach can be used outside the classroom and outside the learning environment entirely for such purposes.

As student recruitment Here, we have an example of a product where the marketing brochure says that by providing market intelligence throughout each phase of the funnel management process. The final that is being the funnel of prospect of incoming students. It's wide at the top which is all the possible prospects that you might gain and narrow at the bottom.

Those prospects who are actually most likely to attend your institution Prospect inquiry applicant accepted deposited registered and matriculant. These are all stages of the funnel management process marketing and recruiting pivots. We read can be made based on changes in student responses and success indices. There's a diagram here on the slide that shows a number of the points where analytics and predictions can be used in order to make this process more efficient and more accurate.

That's all we have for applications of analytics at this time is the short video but I think we like it like that we could probably imagine more applications of predictive analytics. This was just a quick survey of them as we're continuing through this module we're adding additional applications. In other words, additional uses of predictive analytics and coming up with more and more examples of predictive analytics.

Informing learning and teaching technology. That's it for this video. I'm Stephen Downs. Hope you enjoyed it. See you next time.

Force:yes