Business can also use this to make forecasts, anticipate changes, and proactively respond to potential problems. The idea is that with these models, businesses can acquire, grow, and retain customers; plan, manage, and maximize operations; and monitor, detect, and control issues.
Predictive analytics uses historical data in order do draw conclusions about what's happening now and what might happen in the future. Data collected from everything from help-desk chat transcripts to survey results to demographics to payment history can be used to create a predictive model.
The model is built by inputting the historical data, creating a training set and a testing set, applying an algorithm to the data, outputting a model, testing the testing set on the model, and making predictions once the model has been validated. There is also a scoring process involved on the historical data because not all data has predictive power, and some variable actually have overlying causes and effects. This process is not going to accurately predict 100% of the time, but the model can be trained to create something as close to that as possible.
It's pretty apparent how predictive analytics can be used in business, but how is it applicable to education? In many ways, institutions of higher education are like businesses - recruiting and retaining students. There are also a lot of potential problems in schools at all levels including students failing courses, needing to take remedial level work, dropping out of school, or not finding much success in general.
More specifically, schools can use predictive analytics to build early warning systems to predict drop-out-like behavior or chance of graduating high school or college in four years. It can also help predict how prepared students are for college or the workforce. And predictive analytics can also help to identify struggling teachers or indicators of successful teachers. Intelligent tutoring systems might also use this to recommend problem sets or tutoring resources; this also could be used to recommend a series or order of courses for a student.With the prediction models, key stakeholders can get a more clear picture of individuals
So how are these models created? Programs like RapidMiner can be used for predictive modeling. In this program, the user inputs the data, sets the role or roles, and applies an algorithm to create the model.
Regression models are ones in which the "answer" is already known for the training data, such as, based on this historical data, did the student drop out of high school? yes or no. Some types of regression models are used for binary predictions like this yes/no question, and others are used to predict a number, or one of a set of categories (more than two).
Algorithms can also use "if...then" phrasing to make predictions. For examples, if a students spend a very little time on problems and then gets them correct, then the prediction might be different if a student spends a very little time on problems then gets them incorrect; whereas spending a long amount of time on a problem and getting it correct is not the same as spending a little amount of time on a problem and getting it correct.
In education, data can come from many of the same places as business data, but education also provides test scores, engagement levels measured in time, boredom behaviors, correctness, attendance, participation levels, and so on. The key to some types of predictive analytics is discovering which of these features are most effective at creating the predictive model.
It's important that predictive modeling is not the only system in place for intervening with students and problem solving in education. Sometimes data can't tell the whole story. The student should also be invited to be a part of that story. This does have the potential, if used effectively to create systematic change that could help solve many of the problems of our modern-day educational system.
References:
- Rapidminer. (2014). Introducing Predictive Analytics and RapidMiner. Retrieved from https://www.youtube.com/watch?v=ma14K56fNAM&spfreload=10
- Hill, B. (2013). What is predictive analytics? Retrieved from https://www.youtube.com/watch?v=0KXME_y-3QA&spfreload=10
- Rethinam, V. (2014). Predictive analytics in K-12: Advantages, limitations & implementation. THE Journal. Retrieved from http://thejournal.com/articles/2014/06/12/predictive-analytics-in-k-12-advantages-limitations-implementation.aspx
- Baker, R. (2014). San Pedro Case Study. Retrieved from youtube.com/watch?v=TShkg7lZhto
- Essa, A., & Ayad, H. (2012). Improving student success using predictive models and data visualisations. In Reserach in Learning Technology (Vol. 5, pp. 58–70). Retrieved from http://www.researchinlearningtechnology.net/index.php/rlt/article/view/19191
- Baker, R. S. (2014). Chapter 1: Prediction Modeling. In Big Data in Edcuation. New York, NY: Teachers College, Columbia University. Retrieved from http://www.columbia.edu/~rsb2162/bigdataeducation.html
No comments:
Post a Comment