Graded Project
Machine Learning - Unsupervised Learning
Domain:
○ E-commerce
Business Context:
● Customer segmentation is one of the most important marketing tools at your
disposal, because it can help a business to better understand its target audience.
This is because it groups customers based on common characteristics.
● Segmentation can be based on the customer’s habits and lifestyle, in
particular, their buying habits. Different age groups, for example, tend to
spend their money in different ways, so brands need to be aware of who
exactly is buying their product.
● Segmentation also focuses more on the personality of the consumer,
including their opinions, interests, reviews, and rating. Breaking down a
large customer base into more manageable clusters, making it easier to
identify your target audience and launch campaigns and promote the
business to the most relevant people
Dataset Description:
The dataset contains measurements of clothing fit from RentTheRunway.
RentTheRunWay is a unique platform that allows women to rent clothes for
various occasions. The collected data is of several categories. This dataset
contains self-reported fit feedback from customers as well as other side
information like reviews, ratings, product categories, catalog sizes, customers’
measurements (etc.)
Attribute Information:
[Link] Attribute Description
1. user_id a unique id for the customer
2. item_id unique product id
3. weight weight measurement of customer
4. rented for purpose clothing was rented for
5. body type body type of customer
6. review_text review given by the customer
7. size the standardized size of the product
8. rating rating for the product
9. age age of the customer
10. category the category of the product
11. bust size bust measurement of customer
12. height height of the customer
13. review_date date when the review was written
14. fit fit feedback
Data Citation:
● Rishabh Misra, Mengting Wan, Julian McAuley "Decomposing Fit Semantics
for Product Size Recommendation in Metric Spaces". RecSys, 2018.
● Rishabh Misra, Jigyasa Grover "Sculpting Data for ML: The first act of
Machine Learning". 2021.
Project Objective:
Based on the given users and items data of an e-commerce company, segment
the similar user and items into suitable clusters. Analyze the clusters and provide
your insights to help the organization promote their business.
Steps to the project: [Total score: 50 points]
● Import the required libraries and load the data: [ Score: 3 point ]
1. Load the required libraries and read the dataset. (1)
2. Check the first few samples, shape, info of the data and try to familiarize
yourself with different features (2)
● Data cleansing and Exploratory data analysis: [ Score: 20 point ]
3. Check if there are any duplicate records in the dataset? If any, drop them.(1)
4. Drop the columns which you think redundant for the analysis.(Hint: drop
columns like ‘id’, ‘review’) (1)
5. Check the column 'weight', Is there any presence of string data? If yes, remove
the string data and convert to float. (Hint: 'weight' has the suffix as lbs) (2)
6. Check the unique categories for the column 'rented for' and group 'party:
cocktail' category with 'party'. (2)
7. The column 'height' is in feet with a quotation mark, Convert to inches with
float datatype. (3)
8. Check for missing values in each column of the dataset? If it exists, impute
them with appropriate methods. (3)
9. Check the statistical summary for the numerical and categorical columns and
write your findings. (3)
10. Are there outliers present in the column age? If yes, treat them with the
appropriate method. (3)
11. Check the distribution of the different categories in the column 'rented for'
using appropriate plot. (2)
● Data Preparation for model building: [ Score: 2 point ]
12. Encode the categorical variables in the dataset. (1)
13. Standardize the data, so that the values are within a particular range. (1)
● Principal Component Analysis and Clustering: [ Score: 23 point ]
14. Apply PCA on the above dataset and determine the number of PCA
components to be used so that 90-95% of the variance in data is explained by
the same. (7)
15. Apply K-means clustering and segment the data. (You may use original
data or PCA transformed data) (8)
a. Find the optimal K Value using elbow plot for K Means clustering.
b. Build a Kmeans clustering model using the obtained optimal K value
from the elbow plot.
c. Compute silhouette score for evaluating the quality of the K Means
clustering technique.
16. Apply Agglomerative clustering and segment the data. (You may use
original data or PCA transformed data) (8)
a. Find the optimal K Value using dendrogram for Agglomerative clustering.
b. Build a Agglomerative clustering model using the obtained optimal K
value observed from dendrogram.
c. Compute silhouette score for evaluating the quality of the Agglomerative
clustering technique. (Hint: Take a sample of the dataset for agglomerative
clustering to reduce the computational time)
● Conclusion : [ Score: 2 point ]
17. Perform cluster analysis by doing bivariate analysis between cluster labels
and different features and write your conclusion on the results. (2)
Submission:
- Please submit the solution file in .html and .ipynb format on Olympus
- Add necessary comments wherever required.