Jan 2026

Creating a simple tool that uses Machine Learning, APIs, AI & More

Machine Learning Web Scraping/SEO Google APIs AI & NLP

What the app does & why I decided to create it

I wanted to see - just by using a property address, what kind of information I could pull into a web app that would be useful for assessing competitors in the market.

Plus, I thought it would be a fun challenge to build a tool that uses Machine Learning, APIs, etc. as I continue to learn about these systems/technologies and various use cases.

In short, the app does the following:

  1. 1. Retrieves images of the property (street view & aerial)
  2. 2. Pulls in demographic data (zip code)
  3. 3. Creates a map with pins of nearby competitors
  4. 4. Creates a list of competitors with distances, SEO ranking, customer sentiment and more

So, how does this work?

Machine Learning

  1. First TensorFlow/Keras models were pre-trained with self storage imagery (street view and aerial views for Class A, B, C properties).
  2. The python script retrieves the subject property address then processes the photos.
  3. The model outputs the Class and displays a % confidence.
Machine Learning Model Training

Street view images are hit and miss and often have vehicles/off angles so it's hard to really train the model well on this view. The aerials are much better but still there are some limitations depending on Class B/C properties and the subtle differences from the aerial imagery.

Web Scraping/SEO check

  1. For each competitor URL there is a HTTP GET request made and the script uses BeautifulSoup library to check for basic SEO elements.
  2. 1. Page title
  3. 2. Meta description
  4. 3. Mobile viewport tag

This can be challenging with REIT websites (extra space, public storage etc) given the anti-scraping features and dynamically loaded content. So the default if it is a REIT property, there is a high SEO score given.

Machine Learning Classification

Google APIs (Geocode & Places)

  1. Geocoding API converts the subject property address into latitude and longitude coordinates, enabling accurate mapping and spatial analysis.
  2. Google Places API is used to automatically identify nearby self-storage competitors, pull location details, website URLs, and customer reviews.

AI & Natural Language Processing (Customer Sentiment)

  1. Customer reviews pulled from Google are analyzed using natural language processing NLP to estimate overall sentiment.
  2. Reviews are scanned for common themes (pricing, cleanliness, staff behavior, access, etc.), and a concise summary of customer perception is generated.
Machine Learning Results