/Product-Demand-Forecasting-Using-ML

Using machine learning to solve one of the most common problem of Supply Chain domain, i.e Demand Forecasting.

Primary LanguageJupyter Notebook

Demand Forecasting using Machine Learning

Demand forecasting is the process of making estimations about future customer demand over a defined period, using historical data and other information.

Usually organisations follow tranditional forecasting techniques/algorithms such as Auto Arima, Auto Arima, Sarima, Simple moving average and many more.

Table of Contents

Goal

Due to the recent boost in AI world, companies have started researching the possibility of using machine learning in place of tranditional approach.

Tuning traditional algorithms takes a significant amount of effords and domain expertise as well.

In this repo, we are trying to figure out a way of predict the same using machine learning algorithms.

Data

The dataset comprised of units sold on a daily basis along with details regarding the sales, eg. SKU(product id), Store, price etc.

record_ID, week, store_id, sku_id, total_price, base_price, is_featured_sku, is_display_sku, units_sold

Workflow

  • Handling missing values
  • Feature selection based on my previous experience in Supply chain domain
  • Converting dataset into time series format to apply supervised learning approach.
  • Regression Modeling
    • Random Forest
    • XGBoost
    • SVM (future scope)
  • Hyperparameter Tuning

Result

train

Required Packages

  • numpy
  • pandas
  • sklearn
  • easypreprocessing
  • seaborn
  • matplotlib
  • xgboost