Victorian schools trial predictive analytics

Victoria’s Department of Education and Early Childhood Development (DEECD) is considering predictive analytics technology to extract more value from its new enterprise data warehouse.

By the end of next year, the department hopes to conclude a four-year, staged effort to pull together some 180GB of data into its on-premise data warehouse.

That includes Australian Bureau of Statistics data, enrolment figures, test results, teacher, student and parent surveys and financial data of more than 1500 public schools in the state.

To date, DEECD has used its data stores to produce school performance reports for comparison with previous years’ results on a retrospective basis.

The department hopes to develop a forward-looking capability to identify trends and solutions through predictive analytics following a 12-week trial of IBM technology last year.

It hopes to commence a full-scale deployment next year, after addressing data governance and quality issues and undertaking a Privacy Impact Assessment in accordance with Federal recommendations.

Michelle Hill of the DEECD’s data outcome and evaluations division said the technology offered a quick, automated way of identifying and addressing trends like factors that may affect students’ engagement with their schoolwork.

She hoped the technology would allow schools to make evidence-based decisions, instead of having to rely solely on “gut-feel” solutions or lengthy academic studies.

“There was enormous value locked up in segmented data, with no real way of linking it together,” Hill said of the department’s data environment.

“The department could conduct basic analysis of the segmented data, but we didn’t have the capability or tools to conduct advanced data mining across all integrated data sets.

“We realised that by creating an integrated data platform, we could answer more complex queries, for example about the social and economic factors that affect student performance, and based on those facts, make better decisions regarding the strategies that will help schools improve.”

For the trial, which grew from an existing relationship with IBM, DEECD engaged two IBM analysts and one architect to determine where predictive analytics may be applied and what an in-house system might look like.

The pilot drew on data from teacher, parent and student surveys and the standardised Year 7 and 9 examination results of 36,000 students in 300 public schools.

Hill explained that students in Year 9 were at greatest risk of becoming disengaged and dropping out of school.

Results of the pilot confirmed the department’s observations that found that a student’s socio-economic status and that of the entire student body are the strongest predictors of performance.

The IBM consultants worked closely with three DEECD analysts, among the 40 people in Hill’s division.

Hill described the division as a “fairly well-oiled machine” that collected, managed, reported on and analysed data on behalf of the department and its schools, adding some 25GB of new data to the enterprise data warehouse each year.

It would likely undertake the full predictive analytics deployment as an in-house project, she said, noting that its three analysts had gained expertise and a few more staff were “on the cusp” of becoming experts.

“Financial institutions have been using analytics for years, where it’s about building a better understanding of their clients to improve profits,” Hill said. “Our students are our clients.

“The skills transfer process [from IBM to the DEECD] really showed our people what could be done in this area. [Predictive analytics] is a new way of analysing data ... We really want to embed it in our organisation.”