Data Analysis on Local Used Items with Python and Tableau
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Purpose
Why I Chose This Project
I'm fascinated with the second-hand market. Data shows people exchange items in their communities more than ever now. Rachel Botsman's argument in favor of Collaborative Consumption inspires me. I’ve always been partly amazed and partly disturbed by how much we consume produced goods around us.
For better or worse, the internet has hyper-connected us. Boundaries between online and offline life are dwindling in many respects. One of the benefits of this is the ability to buy and sell used goods from local strangers. Not too long ago, most people perceived this as crazy, but we've come a long way.
My hope is that this process will continue to improve and norms of hyper-consumption will re-calibrate into a more balanced state. This is my exploration of local used items where I live in North Carolina.
Questions to Answer
I wanted to answer what locations in my area had the most used items, what the central tendencies of the item prices were by location, and what the number of free items by location was. I also wanted to extract a price and summary of the description of each item, as well as classify each item category using NLP.
Process
Where and How I Extracted the Data
Letgo.com
I used Scrapy to extract out the item information and this proved to be the easiest scrape of all three sites. Letgo.com is JavaScript heavy, but a simple scraper that extracted out the JSON responses was all it took to get what I needed.
https://gist.github.com/Kiwibp/596a77e56144611563934693f9c4d592
Craigslist.org
Craigslist.org proved to be a difficult site to scrape at scale. It has features in place to block IP addresses that it detects as requesting an unusual amount of information in a short period of time.
At first, I tried a simple solution of adding a delay between each page request in my script. But my IP address was blocked after several hundred items scraped. I began to get the impression that Craigslist monitors both the speed of requests as well as the page depth of requests.
In any case, after some research and advice, I created an account on scrapinghub.com and used Crawlera to avoid my IP address getting blocked again.
https://gist.github.com/Kiwibp/e1e2b0f990ad1f5439abc4be76f9aa8c
Facebook.com
Facebook.com is a robust JavaScript heavy website. Unfortunately, I was unable to isolate the JSON requests from Facebook Marketplace to extract the data I needed.
After some research and advice, I determined the simplest approach was to use Selenium. The benefit of using Selenium is you can code any interaction that a user performs on a website. The downside is it scrapes much slower than Scrapy does.
https://gist.github.com/Kiwibp/78cf224a0a5d0c2c33fdb371b8ebdb93
Storing the Data in MongoDB
I imported the JSON objects into Pandas dataframes and the majority of time I spent of the project was cleaning the data. The same primary categories were scraped, but each site had its idiosyncrasies. A good deal of time and effort was necessary to clean the dataframes so they could be merged and produce insights during exploratory data analysis.
https://gist.github.com/Kiwibp/fc3c10f0c0a459184b467c817d5753db
How I Visualized The Data
EDA With Tableau
If you'd like to view the entire Tableau workbook and storyline.
EDA With Pandas & Matplotlib
Results
Insights Gleaned
I had a couple unsuccessful attempts at applying unsupervised NLP with spaCy and pyLDAvis libraries inspired by this walkthrough https://github.com/skipgram/modern-nlp-in-python as well as creating a text summarizer with the Keras library inspired by this walkthrough https://github.com/llSourcell/How_to_make_a_text_summarizer. Then, I decided to simplify the process and use the MonkeyLearn API to execute a text summarizer model as well as a price extractor model. I also created a custom category classification model.
https://gist.github.com/Kiwibp/dac2f4a5bf45155e9267b7cf0d4289bc
Improvements to be Made
I found this project engaging and challenging. If I scrape items again, I would also scrape the designated item type categories. This would make for a more interesting analysis of items by the designated types and I could use them as targets for my NLP classification model.
It appeared that some of the descriptions for the Facebook items weren’t scraped. I wasn’t able to determine why and I would pay more attention to that in the future. The free version of MonkeyLearn only allows 300 queries per month. I would get my customized category classifier more accurate when my allowable query amount resets each month. I would also train it with many more items to see to see if that makes it more accurate.
You can view the Tableau workbook here: https://public.tableau.com/profile/keenan.burke.pitts#!/vizhome/NYCDSAFinalProject_0/LocalUsedItemsAnalysis and my github repo here: https://github.com/Kiwibp/NYC-DSA-Bootcamp--Final-Project.
About Author
