COVID-19 is dramatically changing markets for economic growth. The data used for making good managerial decisions have been turned upside down in this unpredictable marketplace.
Because of the volatility of the situation, all cycle times for reporting were dramatically compressed. In some companies, data teams were made to keep specific pain points under their radar.
Demand forecasting is one of the most difficult challenges for data scientists, even in normal times. As the pandemic hit, shifts in demand have wreaked havoc on ML models that were already slow to adapt to the data which isn’t normally valid.
Should Previously hold data before the pandemic be deleted? Should it be replaced with recent data based on events prior to COVID-19? However, will pre-COVID-19 data stay relevant going forward? The answer to these questions, will obviously, vary by sector. Using moving averages and other smoothing forecasting techniques as a way to navigate how much to rely on pre-and post-pandemic data.
Several analytics leaders have suggested the need to keep eye on newer machine learning models. They plan to audit data input, model assumptions, and model output more frequently. How will models respond to no demand, rapidly increasing demand, or discrepancies like the negative price of oil? Techniques developed for quality control in industrial engineering, like control limits and acceptance sampling, need to be applied to machine learning to make sure the models are in control.
Originally published at https://www.jtpcindia.com.
