Future Sales Predictions on Russian Electronics Shops


  • Jacob Sweeton
  • Daniel Simpson
  • Daniel Roberts
  • Edwards Mantsevich


Speculations about future sales for a given company or store and for given products in those stores have been critical to successful trade going far back in human history. However, until much more recently this was more art than science as the availability of mass sales data was much more limited. Today, we have access to nearly boundless amounts of relevant data and incredible computational tools with which to work with it and to improve upon these speculations. Given the daily historical sales data provided by the Russian firm 1C used in the Kaggle challenge “Predict Future Sales,” we seek to improve on existing sales predictions across items and stores that may or may not be chains. This dataset provides information about many individual store locations as well as 11 fields for products and sales per shop per day from January 2013 to October 2015. To this end, we will use the statistical software R and possibly machine learning methods to generate a month’s worth of sales predictions ahead of store restocking. We will identify common patterns across stores such as sales data for the same or similar product as well as attempting to identify yearly trends in sales. As it is a Russian dataset, we will seek to overcome lingual and cultural barriers faced by international data scientists in industry as we pursue this goal.






Computer Science