The goal of this proect is to build a Realtime multivariate sales forecasting model using deep learning techniques which can be deployed into production using MLOps infrastructure. This practice will help C-Level executives to initiate business strategies.
In this repo we will build a multivariate time series model using different machine/deep learning techniques to forecast multiple products in different stores.
The data used here is taken from Kaggle's Store Item Demand Forecasting Challenge. Below you see a snap shot example of sales of item2 from store 1.
This is a self-supervised learning technique that can learn a compact representation of data. In this case LSTM network is organized into an encoder-decoder architecture which takes an input sequnce and encoded into a context vector (hidden and cell states). The decoder then takes this context vector as an input and produces an output sequence
