Working in manufacturing industry means that you must deal with product failures. As a BI and/or Data Scientist developer, your task is not only monitor and report product’s health state during its lifecycle, but also predict the likelihood of a fail in the production phase or when product has been delivered to the customer. Machine Learning techniques can help us to accomplish this task. Starting from past failure data, we can build up a predictive model to forecast the likelihood for a product to fail or giving an estimate on its duration. And now it is possible to develop an end-to-end solution in SQL Server, because of the introduction of advanced analytics tools like R since release 2016.
In this session, we start from the real case of a manufacturing company to create some predictive models:
a) Regression model, to predict how many months a product will last once it is delivered to the customer; Time to Failure (TTF);
b) Classification models, to predict if a product will fail within a given time frame. (Binary or multi-class classification problem)
The solution is built using SQL Server 2016, R and R services. On top of that, some reports are created to deliver the outcome to the stakeholders.
