shaktipanda1235 / store_footfall_prediction_tsf

Predicting footfall of customers in a given store using SARIMAX model where exogenous predictors helped in better prediction.

Inspiration

Beyond Machine Learning Models, Time Series Forecasting plays an integral role for bussiness depending upon outlets and sales. So I tried an approch of solving a most common day to day customer visits to outlets, whose prediction can be bettered using both Linear Regression and Time Series Forecasting.

Explratory Data Analysis

• About 20-25% of the values are zero for customers which may be an indication of store closure for that day.

• Impact of the promotion is clearly visible in the grouped boxplots below even across month. However for December, no impact of promotion is visible.

• Customer count was high when there is a school holiday.

  

Model Building

• Dickey-Fulher tests suggested stationarity of time-series.

• Various models are built with and without considering seasonality and a SARIMAX model with Time Varying Linear Model performed the best according to rmse values.