Amazon Hackathon Challenge 2022
Problem Statement:
Given online sales data of Dog Food generate a process to evaluate a "WOOF" score - a numeric output of a function or model that will indicate the appropriateness in advertising a dog food brand with epect to other selected features.
Project Summary:
1) Exploratory Data Analysis
2) Predictive Modeling (Agglomerative Clustering, K-means Clustering)
3) Evaluation, Analysis & Insights
Project Detail:
The goal of this project is to help a dog food brand that currently has only one product with their new product strategy: expand their product line with a new product that targets exisiting customers with a similar price range as their exisiting product.
We take the clusters generated using our unsupervised learning models (Agglomerative Clustering and K-means Clustering) to caculate the "WOOF" score. Since the goal is to maximize profits, we calculate the revenue, i.e. the average of price times quantity for each cluster and rank the clusters based on the average revenue of each cluster. Lastly, we nornalize the revenue to between 0 and 1 and call this the "WOOF" score.
With our "WOOF" score, we are able to identify the best performing brand - the brand which has the highest occurrence in the most-valuable cluster which generates the highest average revenue. In addition, we can develop a product strategy based on their product in the similar price range and develop a product strategy accordingly.
Presentation Video