Demand-Forecasting

• Business Problem

A chain of stores, 10 different stores and 50 different requests a 3-month demand forecast for its product.

• Dataset Story

This dataset is presented to test different time series techniques.

Information of 10 different stores and 50 different products in 5-year data of a chain of stores is located.

• Variables

▪ date – Date of sales data

~No holiday effects or store closures

▪ Store – Store ID

~Unique number for each store.

▪ Item - Item ID

Unique number for each product.

▪ Sales – Number of products sold,

~The number of products sold from a particular store on a given date

• Task

• Relevant store using the following time series and machine learning techniques

• Create a 3-month demand forecasting model for the chain.

▪ Random Noise

▪ Lag/Shifted Features

▪ Rolling Mean Features

▪ Exponentially Weighted Mean Features

▪ Custom Cost Function (SMAPE)

▪ Model Validation with LightGBM