The meeting is an inaugural session and will be a working Pizza Dinner.
- Meet and Greet - 10 min.
- About the Group - 5 min.
- Topic of the day - 40 min.
- Q&A, Conclusion - 5 min.
Chris Jackson will share a BigQuery/ML solution for those who need to make decisions about inventory levels using data from past customer behavior to make predictions about likely future purchases. We will see how to use BigQuery ML SQL statements to train and deploy a demand forecasting solution.
BigQuery ML allows you to train multiple time series models, one for every item in your data, in a single query to forecast all products simultaneously. Behind the scenes, an autoregressive integrated moving average (ARIMA) is used to build and test up to 42 models for each item’s time series. BigQuery will automatically return only the best model, with the lowest Akaike information criterion (AIC), for each of your products. Pre-processing of model data includes automatic cleaning adjustments to address missing values, duplicated timestamps, and spike anomalies. And with a minimum of one year of daily data, you can specify a region to account for holiday effects.
About the Speaker:
Chris Jackson is a Cloud Engineer at Google and works with many large scale enterprise companies using Google Cloud in Indiana, Ohio, and Kentucky. Chris specializes in cloud infrastructure but enjoys learning and working on various ML solutions. On a personal level Chris lives on the west side of Indianapolis, has 9 boys (who are all just as nerdy as himself), and is currently working on getting his next level up Ham radio license and learning Morse Code (the original digital mode).