My university has a building on campus where electrical usage and water usage is monitored down to the second. Many electrical loads and water usageis tracked and stored in a database. In the water dataset, the column headers are the location, and the rows contain the gallons used (per second). In the electrical dataset, the power consumption (in Watts) is tracked in the rows and the load type is the column header.

My general problem is as follows: - how could I use the data from this highly monitored building, to benefit the university and give them insights into buildings that aren't monitored like this one?

My data is of the time-series, so I am trying to think of certain clustering or classification algorithms that could help me extract actionable data from it, and perhaps even association rule learning to find relations between the water and electricity data?

Thanks for any insights!


In order to draw conclusions from this one building's data on other buildings on the campus, you would need a set of parameters shared between the building you know about and the buildings you want to predict for. For example, if the monitored building is of a male dormitory type, perhaps you can infer something for male dormitories; if the building is at location X, perhaps you can infer something for buildings at or close to location X. You'd need as many as possible of such additional columns to extrapolate knowledge to other buildings.

As for visualizing/predicting resource usage of that single building, I suggest you try data mining tool Orange, which I find phenomenal for prototyping machine learning workflows. It also has a time series add-on, which I can't vouch for, but it does seem to have some forecast capabilities. Forecasting using VAR model in Orange


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.