Case Study: Machine learning improves data quality and forecasts
In one of our projects with an infrastructure provider, it quickly became clear how much the combination of different data science models can improve business processes. In the industrial park, toxic wastewater is transferred to a sewage treatment plant via a sewer system. In order to safeguard the water quality, counteractions must be initiated as quickly as possible in the event of irregularities. Regular samples of water temperature, quantity, conductivity or pH value are taken by sensors. With the help of critical anomaly detection, it was possible to directly identify data outliers..
After our data scientists were introduced to the company’s complex data structures and operational processes, they developed an algorithm as part of a proof-of-concept that alerts in the event of irregularities. With the help of Unsupervised Learning, outliers in data quality could first be identified and then, based on this, it was possible to predict when a worrying wastewater value will occur using target values with the help of Supervised Learning.