Designing a product recommendation system based on taxonomy

As online shopping continues to gain popularity, personalized recommendations have gained significance in e-commerce. Finding exactly what a customer wants might be difficult due to the millions of goods that are accessible online. This is where personalized recommendations can help by giving users recommendations that are specific to their needs and habits. Taxonomy-based recommendation systems are one method for individualized suggestions. It is simpler to search for and retrieve information when objects or concepts are organized and classified according to a taxonomy, which is a hierarchical structure. In this article, we'll take a closer look at a taxonomy-based product recommendation system and how you can develop one on your own.

What is taxonomy?

Taxonomy is an approach for categorizing and organizing things into hierarchical structures. To make it simpler for people to search for and locate what they're searching for in e-commerce, the taxonomy can be used to classify and subgroup items. Systems for recommending content that is taxonomy-based provide a number of benefits over conventional recommendation systems. Accuracy improvement is a significant benefit. suggestions can be produced based on how similar the items in a category are to one another, increasing the relevancy of the suggestions, by classifying products according to their attributes. A further benefit is enhanced user experience. Users are more likely to locate items that match their needs and are more inclined to visit the website when given personalized suggestions that are catered to their interests and preferences. enormous and diversified datasets can be handled by taxonomy-based recommendation systems, making them appropriate for e-commerce platforms with enormous inventories.

How does taxonomy work?

Taxonomy-based recommendation systems classify items based on their traits into groups and subgroups. Depending on its characteristics, such as brand, price, color, or size, each product is categorized into one or more groups. When a user searches for or sees a certain product, the system makes recommendations for further goods in the same category or subcategory based on their behavior and interests.

For instance, the algorithm can suggest alternative gowns in the same color or design if a user searches for a blue dress. Several algorithms, including the KNN (k-nearest neighbors) algorithm, which calculates the distance between products based on their features, are used to determine how similar the products within the categories are to one another. Businesses can boost customer happiness and loyalty by employing taxonomy-based recommendation systems to offer consumers personalized recommendations that are pertinent to their interests.

Designing a product recommendation system based on taxonomy

We will utilize the pandas library for data processing and analysis as well as the scikit-learn package for machine learning strategies while developing a taxonomy-based product recommendation system. To explicitly illustrate the taxonomy-based recommendation system, we are going to create this dataset.

The KNN (k-nearest neighbors) method, a machine-learning technique that determines the distance between data points based on their attributes, was used to construct the recommendation system. The goal of this research is to demonstrate how taxonomy-based recommendation systems can be used to offer people personalized recommendations based on their interests and behaviors.

Firstly, let's import the necessary libraries −

import pandas as pd
import numpy as np
from sklearn.neighbors import NearestNeighbors

Let's next build a dataset for products using our taxonomy:

products = pd.DataFrame({
   'product_id': [1, 2, 3, 4, 5, 6, 7, 8, 9, 10],
   'category': ['Tops', 'Tops', 'Tops', 'Bottoms', 'Bottoms', 'Shoes', 'Shoes', 'Accessories', 'Accessories', 'Accessories'],
   'sub_category': ['T-Shirts', 'Shirts', 'Sweaters', 'Pants', 'Jeans', 'Sneakers', 'Boots', 'Jewelry', 'Hats', 'Bags'],
   'material': ['Cotton', 'Cotton', 'Wool', 'Cotton', 'Denim', 'Leather', 'Leather', 'Gold', 'Cotton', 'Leather'],
   'style': ['Casual', 'Formal', 'Casual', 'Casual', 'Casual', 'Casual', 'Formal', 'Formal', 'Casual', 'Casual'],
   'color': ['White', 'Blue', 'Gray', 'Black', 'Blue', 'White', 'Black', 'Gold', 'Red', 'Brown'],
   'size': ['S', 'M', 'L', 'S', 'M', '10', '11', 'NA', 'NA', 'NA'],
   'brand': ['Nike', 'Ralph Lauren', 'Tommy Hilfiger', 'Levi's', 'Wrangler', 'Adidas', 'Steve Madden', 'Tiffany', 'New Era',      'Coach']
})

In the process, a dataset of ten products is produced, each of which has a category, subcategory, material, style, color, size, and brand.

Following that, we must transform the categorical data into numerical data. To turn the category information into numerical information, we'll employ one-hot encoding.

products_encoded = pd.get_dummies(products[['category', 'sub_category', 'material', 'style', 'color', 'size', 'brand']])

After encoding our data, we can now fit a KNN model to our dataset

knn_model = NearestNeighbors(metric='cosine', algorithm='brute')
knn_model.fit(products_encoded)

Using the KNN model to discover the K nearest neighbors to a given product, we can now make a prediction −

def get_recommendations(product_id, K):
   product_index = products[products['product_id'] == product_id].index[0]
   distances, indices = knn_model.kneighbors(products_encoded.iloc[product_index, :].values.reshape(1, -1), n_neighbors=K+1)
   recommended_products = []
   for i in range(1, K+1):
      recommended_products.append(products.iloc[indices.flatten()[i], 0])
   return recommended_products

And finally, we make the predictions with the following code.

print(get_recommendations(1, 3))

Output

[4, 3, 9]

In this example, we constructed a dataset of ten goods and used one-hot encoding to transform categorical data into numerical data. We then fitted a KNN model to the dataset and built a function to provide suggestions based on the K nearest neighbors to a particular product. Finally, we tested the recommendation system by obtaining the top three suggestions for product_id 1.

Conclusion

Finally, a taxonomy-based recommend-er system is a great way to build a product suggestion system. The system can give personalized suggestions to clients by categorizing things based on their attributes, enhancing their platform experience. However, creating an effective taxonomy and gathering product data can be difficult, and recommend-er systems may struggle with the cold start problem and scalability. In the end, a taxonomy-based recommend-er system is a powerful tool for e-commerce businesses looking to enhance user experience, boost revenue, and minimize attrition.