skrypnyuk/grupo-bimbo-inventory-demand

Grupo Bimbo Inventory Demand

This repository contains my project code and project report for the Grupo Bimbo Inventory Demand Kaggle competition. The goal of the project was to develop a model that could most accurately forecast inventory demand for Grupo Bimbo products baed on historical sales data.

Competition Information and Data

All information and data can be obtained from the competition web page:

https://www.kaggle.com/c/grupo-bimbo-inventory-demand

iPython Notebooks

1. exploratory analysis:

* Explores product and client data.
* Discovers new features that can be generated from NombreProducto 	and NombreCliente variables.

2. feature engineering:

* Takes the findings from the exploratory analysis and creates new 	product and client data sets with additional features.

3. build:

* Load train, test, and modified client and products data.
* Merges modified client and products data into train and test 	data.
* Adds time series demand features.
* Adds mean of frequencies of id features.
* Encodes categorical variables.
* Removes data before Week 6.
* Writes modified train and test data to CSV.

4. predict:

* Fits an XGBOOST model to the train data.
* Validate model results on a held-out subset of the train data.
* Generates results for the test data.

5. free-form-visualization:

* Uses Matplotlib to plot weekly sales of the top 4 best-selling 	products in the train data. 

Project Report

This is the PDF report for the Machine Learning Engineer Capstone project. It details the process I implemented, the analyses performed, and the models built to solve the goal of the Grupo Bimbo competition.

References and Sources

1. XGBOOST: https://github.com/dmlc/xgboost

2. Sklearn: http://scikit-learn.org/stable/

3. Owen Zhang's Tips for Data Science Competitions: http://www.slideshare.net/OwenZhang2/tips-for-data-science-competitions

4. Classifying Client Type using Client Names: https://www.kaggle.com/abbysobh/grupo-bimbo-inventory-demand/classifying-client-type-using-client-names

5. Exploring Products: https://www.kaggle.com/vykhand/grupo-bimbo-inventory-demand/exploring-products