When a business is setting up its e-commerce website for the first time without any product ratings and user purchase history needs a recommendation system. We have made a search based recommendation system that recommends products based on the product descriptions of the products. Based on the words present in the product descriptions, this system makes clusters of words associated with similar products. Based on these clusters, products can be recommended based on the search words entered by the user. Our model uses k-means clustering in order to make these recommendations. The dots on the graph represents the words present in our cluster. These words are segregated into a hundred clusters for this particular sample dataset.
We have made this demo for a sample dataset that can be downloaded. Based on the sample products present in the drop-down menu, words present in the descriptions of the recommended products are displayed in the output box.