Dash, Recurring Neural Networks (RNN), and SHAP for Predictive Maintenance

Recorded on September 15, 2021

With modern supply chain models, it is critical for businesses to ensure the undisrupted functionality of their machinery and equipment. To monitor and anticipate potential anomalies, companies turn to predictive maintenance.

Plotly’s Dash and Vaex deliver predictive maintenance dashboards in a simple, user-friendly way. Data scientists can easily process Big Data using Vaex and add models with open-source Python libraries like Keras and SHAP.

In this webinar, Vaex co-Founder, Jovan Veljanoski, will use data from the NASA Turbofan Degradation Simulation to demonstrate how easy it is to build, deploy, and scale fully interactive web apps. No JavaScript or DevOps required. He will walk through the following:

  • Simplifying data preprocessing with Vaex's powerful platform
  • Building a Recurrent Neural Network model using Keras
  • Improving model interpretability via SHAP
  • Productionizing the entire model pipeline by using Dash and Vaex together

About Jovan Veljanoski

Jovan is a senior data scientist & researcher at Cloud Technology Solutions, where he creates predictive models and data pipelines for a variety of domains and use cases. Working mostly with Python in the Jupyter/PyData ecosystem, he has considerable experience in creating dashboards, clustering analysis and predictive modeling. Jovan has a PhD in Astrophysics, is a co-founder of vaex.io, and is interested in novel machine learning technologies and applications.