Inspiration

I work for AySA, the Argentinian public company dedicated to supplying millions of Argentinians with running water and sewer services, which in the last few years has been giving more room for innovation.

The encouragement to participate in the Fiware4water challenge is a proof of that, and we saw it as an opportunity to learn useful and innovative things and as a space to think and share ideas related to the field.

What's the project about

It contains the results of validating sensor data and grouping wastewater by the concentration of different substances contained within it based on CAP data

How we did it

We calculated the correlations using the Pearson's coefficient since we decided to avoid rank correlations (mainly due to the huge number of ties), and we calculated the clusters with k-means after dealing with the outliers. There was a small number of vectors therefore we thought it was better to remove the outliers "manually" instead of trusting an algorithm to do it.

Challenges we ran into

We had a dilemma regarding whether to use sliding windows of time. Since in the provided data there were different intervals between days, we decided not to use that paradigm.

Built With

• python
• rapidminer