Predicting Liquor Sales
Machine learning has amazing potential to completely rewrite some of the business procedures that originally, were not far from guessing. Machine learning takes the understanding of statistics and a business's' historical growth and combines it with the raw computing power of machines. This marriage has led to unprecedented understanding by creating sales models with superb accuracy.
To research how machine learning in sales could lead to better projections models, I applied regression and regularization techniques to state liquor sales by zip code and store number using publicly available data from the State of Iowa's website. I used data from 2015 to estimate total sales for each store in 2016, extrapolating from the sales for the first quarter of 2016. This data helped create models for 2016 with stunning accuracy, some above 98%.
Complete write up and code available on my GitHub.