Chemical Composition Analysis for Skincare Products

Abstract

This project explores the use of t-Distributed Stochastic Neighbour Embedding (t-SNE) to visualize and analyze moisturizer products based on their ingredient compositions, specifically focusing on those designed for dry skin. The dataset includes detailed product labels, brands, prices, full lists of ingredients, and information about their suitability for different skin types. Ingredient information is transformed into a structured numerical representation using one-hot encoding, forming a document-term matrix (DTM) that maps ingredient presence across a wide range of products. Given the high-dimensional and sparse nature of ingredient data, t-SNE is applied to effectively reduce dimensionality, enabling intuitive two-dimensional visualization. This process successfully clusters products with similar ingredient profiles, revealing hidden patterns and relationships among them. To make the analysis more interactive and user-friendly, the project incorporates dynamic Bokeh plots. These plots feature hover interactions that allow users to view detailed product information, such as brand name, price, and key ingredients, simply by moving the cursor over a data point. Colour coding is also applied to distinguish different clusters or patterns based on formulation similarities, price ranges, or brand categories. This visualization approach benefits consumers by helping them select suitable moisturizer products that best match their skin needs, particularly those suffering from dry skin. It also assists skincare professionals in identifying formulation trends, such as popular combinations of hydrating ingredients. Furthermore, manufacturers and product developers can leverage these insights to optimize ingredient compositions, create targeted formulations, and improve market offerings. By integrating the dimensionality reduction power of t-SNE with the interactive capabilities of Bokeh, this project offers a comprehensive, engaging, and efficient method for analyzing and understanding moisturizer formulations at scale.