By: Morten Aas-Lyngby, Ignacio Alonso, Paolo Brambillasca, Alberto Dona & Emilio Rodriguez
This repo contains the notebooks to execute the same analyses for our BCG X Capstone.
- Churn Modelling, RFM Segmentation, Risk and CLV Calculation, SARIMA Stock Demand Prediction
Through this repository, you will find 5 official notebooks:
-
BCG X Preprocessing & EDA Notebook: Dives into data preprocessing, missing values inputation, Exploratory Data Analysis and Quantiles Analysis
-
BCG X Churn Modelling: Constructs a churn model based on recency, max purchase lag and median days between orders. It explores models as well as evaluation metrics. It finishes by taking a last look at how it could measure monetary risk.
-
BCG X Product Analysis & Cross-selling strategies: We examine the best selling products with basket analysis and we suggest association rules based on that, not too niche and not too wide. We will look at the metrics such as support, confidence and lift. Finally, we make an economic estimate on the sales improvement.
-
BCG X RFM & Client CLV: Implementing the RFM framework in a weighted manner, giving more importance to the economic indicator. Calculating CLV based on each cluster and subsequently, each client.
-
BCG X SARIMA Model: For Stock prediction, we analyse branches and product demand in a weekly and monthly basis. We achieve significant performance.
Additional Notebooks: We explore other machine learning models (mainly boosting) and consolidate some analysis together to show the power of a combined approach based on customer segmentation, churn probability and associated monetary risk of losing that customer.
- Step 1 - notebooks/: Download the notebooks to your machine.
- Step 2 - data/: create a data/ folder within the folder where you placed the notebooks. Toss the datasets there. (You may not see them without permissions, since we signed an NDA)