Sales observation from Feb'13 to Aug'15 and prediction for Sept-Oct'15
Sales trend throughout yearPredicting future sales of a company is not an easy task, simply because of countless non-linear factors associated. But with recent development in AI / ML, this forecast is possible! 'What is going to be demand for coming quarter?' or 'How much will I be able to sell this year?', precise answer to these questions can help us in better management of warehouse, increase in customer satisfaction and ultimately, brand loyalty among them. This project aims to predict the individual shop's sell for coming specified duration by analysing previous customers' data from around a thousand stores over last 6 months using Facebook's prophet.
- Dataset: (1). Customer Data (~10L) (2). Shop Data (~1100)
- Features: StoreId, NumberOfCustomers, Promo, StateHoliday, SchoolHoliday, StoreType, CompetitionDistance, PromoInterval and 10 others
- Target: Sales/day
- Libraries: Numpy, pandas, seaborn, matplotlib, datetime & fbprophet
- Training: Facebook Prophet
- Insights:
- Overall sale/year has been decreasing since 2013.
- Maximum sale is during holidays, especially Christmas
- Follow below link for complete step-by-step analysis (link will be activated soon)
Github Link
Future Scope
Here, Prophet has been choosen as time series forecasting procedure due to its good fit with seasonal and holiday effect. Other forecast techniques can be explored and with more accurate sales data, Confidence Interval can be further improved.