- Check chosen factors on customers quantity and cart value
- Store type
- School-free days
- Special Offer
- Work days and week days
- Month splitted by 3 parts
- Month splitted by weeks
- Weather
- Seasons
- Months
- States
- States position
- Others
- Use regression from ols model and A/B test.
- You can use unsupervised learning for searching focus groups
- Try to define what are types of shops (a, b, c, d)
- check factors influence on sales
- regression
- A/B tests
- Try to propose a method of analysis that will allow us to determine what store types a, b, c, d might mean.
- build model for every day of week decoupled from season and global factors
- find anomalies and their causes
- estimate mean sales for next 8 weeks (June and July) based on data to May 2015.
- designate model to data matching error
This is only summary. Full description in notebook
- Functions for generic operations
- Heatmap of greater part of data was created
- Influence on sales, customers quantity and cart value was checked on factors like:
- store type
- assortment
- special offer
- school-free days
- days of the week
- month phase
- weather
- storm
- rain
- thunderstorm
- fog
- hail
- months
- states
- also was checked dependence between:
- store types and assortment
- hail and assortment
- states and store types
- states and assortment
- states and school-free days
- Depending on sales, factors such as store type, assortment, promotion, phase of the month, month, hail, day of the week were used.
- Sales prediction with 0.769 R-Squared mapping compared to real data on whole time range.
- Sales prediction with 0.999 R-Squared mapping compared to real data on next 6 month.
- Adding the State/StateName factor significantly reduces the result value.