Product Matching Using Machine Learning
What does the product look like?
The first step of any ML project is to identify what features we want our model to focus on. We use these features to train our machine learning algorithm. Features can be anything that helps us understand the data we have gathered. In this case, we will be using the image of the product to determine if we should match it with a specific product.
How do I gather my data?
For this project, we will be gathering product images and their corresponding labels. These images will be fed into our model and then trained to recognize patterns in the data. Once we have enough training data, we can start building our model.
1. Amazon product search
Amazon’s product search page is pretty simple. You enter some keywords related to what you’re looking for and it returns a list of results. If you click on any result, it takes you to the product details page. On the left-hand side, you’ll see three tabs: ‘All’, ‘Similar’, and ‘Customers who bought this item also bought. Clicking on these three tabs shows you the top 10 items based on how closely they match the keyword you entered.
2. Similar
Click on the similar tab, and you’ll get a list of products that look like yours. These are the ones that are closest matches to your original query.
3. Customers who bought this item also purchased
This is where you’ll find out if anyone else has bought the same thing. If someone else has bought the same product, then it means that the two products are probably a good fit.
4. Customers who viewed this item also viewed
If you click on this tab, you’ll get a list of people who looked at this particular product. If you click on a person’s name, you’ll go straight to their profile page.
5. Products that customers bought together
You can use this feature to see if anyone else has bought a combination of products that you’ve got. So if you’ve got a dresser, bedside table, and lamp, you could check if anyone else has bought those products together.
6. Related searches
This is where you can see what people searched for after searching for something similar to your product. If you click a link, you’ll go to a page showing the top 10 results for that search term.
What type of model am I going to use?
In order to make sure we get good results, we will need to choose a good model. There is no single best algorithm for everything, so we will need to find the right balance between accuracy and speed. We will use convolutional neural networks (CNN) to build our model. CNN’s are great at identifying patterns in data.
How do I train my model?
Once we have chosen our model, we need to figure out how to train it. We will use Tensorflow to do this.
How do I test my model?
We will be testing our product matching model by feeding it a set of products and seeing how well it matches them. If we get a high percentage of correct matches, we will know that our model worked!
