Industrial Machine Learning

Data scientists developing a predictive maintenance service at one industrial sensing company were faced with two problems: how to interactively sift through their raw data; and how to measure their model effectiveness over time. In particular, they wanted to compare the performance of various model iterations to measure improvement.

With TimescaleDB, not only are they now able to store their raw data, but also capture metadata related to the model’s performance: accuracy, latency, etc. Because TimescaleDB supports pure SQL, they use their existing tools to query the data. Our complex query support means they can monitor model effectiveness by cohort, providing a real-time machine learning dashboard.