How to improve a Marketplace: Project Overview

Segmented sellers based on sale/no_sale as well as on avg LTV (rmse ~BRL289) to find what characteristics makes a top seller, following CRISP-DM process model.

• Business Understanding
• Data Understanding: Explored and Wranged data from 9 datasets with +100k observations (one of the datasets have +1M observations!) and max 15 variables. Wordclouds of good/bad/neutral scored reviews
• Data Preparation: Engineered Features and cleaned city names. Dealt with missing values and encoded categorical data.
• Data Modeling: Optimized Random Forest Regressor and Classifier using GridCV.
• Results Evaluation: Extracted actionable insights from model with Partial Dependant Plots.

Code and Resources Used

• Python Version: 3.8.5
• Packages: pandas, numpy, sklearn, matplotlib, seaborn, plotly, fuzzywuzzy, wordcloud, graphviz, pdpbox

Repo Walk-Through

Please read this project' post for a business point of view

• 00byMontse.py: A few auxiliary basic functions. i.e. Summary data function based on describe() method
• 01Olist_EDA: Explored data on each dataset individually to better understand the provided data. Looked at the distributions of the data and value counts.
• 02Olist_DataWrangling: Merged, Cleaned & Prepared Data to create a 'basic Seller Dataset' based on their characteristics, acquisition journey and sales to date. Also wranged a second version including (cleaned) location data.
• 03Olist_ProductBusinessSegments: Answered business questions: -Is there any Business Segment/Product Category doing particularly well … or particularly badly? (Sellers acquired between Dic,2017-Aug,2018)
• 04Olist_SellerSegmentation_Binary: Seller Segmentation based on Sale/NoSale. What does it take to sell?
• 05Olist_SellerSegmentation_LTV: Seller Segmentation based on avg LTV (1 month). What does it take to be a top seller?
• 06Olist Channels:What Channels bring in the most Top Sellers?

Acknowledgements (and praise!):

• JungJoon Lee - on EDA with eCommerce Marketplace (Seller Side)
• Eduardo Cuducos - on getting a clean list of Brazillian cities
• Gene Diaz jr - on stopwords in Portuguese
• Sankarshana Kadambari - on how to clean city column