Demand Forecasting with Data Science

Welcome to Part 2 of the demand forecasting in IIoT blog series. In this blog we will talk about the importance of utilizing sensor data to improve demand forecasting model. The blog also explores the questions to consider when choosing the appropriate sensors as the model input. Be sure to check out parts 1, 3, and 4 in the series as well.

Utilizing Sensor Data

It is common for industrial IoT costumers to utilize sensors to monitor their systems and operations. The sensor data has been primarily used by line managers or technicians to manage their daily operations. Depending on the type of sensor data, some of them could be utilized for forecasting demand. For example, a security camera at the entrance of a restaurant can provide insights into the restaurant traffic and thereby help in forecasting food sales. Another example would be utilizing weather and solar panel reliability data to predict the energy output. Since typically the sensors are not configured for forecasting demand, it is important to differentiate signal from noise. The quality and validity of the sensor data is an important factor when considering relevant sensor types. Below are some questions that will help validate the sensor data quality:

  1. Can the sensor data possibly influence the demand?
  2. Is the sensor data representative of the information?
  3. Does the sensor data have any missing or null values?
  4. What is the frequency of transmission for the sensor data?
  5. Does the sensor data have default values?
  6. Was there a test period for the sensor? If yes, are the sensor values valid during that period?

Exploration of the above questions along with analysis of the sensor data will help determine the best sensor types and time periods to choose for demand forecasting model. Once the appropriate sensor data is chosen, then it is important to link the sensor data with the demand data to enable forecasting. If an exact time for linkage cannot be determined, then it is worth checking for time windows to join the two datasets. The time windows can be determined by interviewing the customer or by manually checking the time difference between the sensor trigger and the actual demand time.

Stay tuned for our next blog in this series.

Are you an industrial IoT customer looking to save money by having an efficient demand forecasting model, then talk to us. Our 20 years of experience in IoT combined with our expertise in data science, will help you navigate the complex world of demand and customer behavior.  For questions about how Bsquare can help reduce your operational cost, talk to us.