Product Recommendation and Product Matching Using Deep Learning
Author : Dr. Gandhimathi K, Savitha B and Shrinithi V
Abstract :
The swift expansion of e-commerce has rendered product selection increasingly intricate, influenced by factors such as pricing, customer feedback, and ratings. This project introduces a recommendation system based on deep learning, aimed at aiding users in identifying the most appropriate product within a specified category. By utilizing structured data, which includes price, ratings, and sales channels, alongside unstructured customer reviews, the system enhances the precision of its recommendations. The dataset is subjected to preprocessing, where categorical variables are transformed through label encoding, and numerical features are standardized using StandardScaler. Customer reviews are processed through tokenization and converted into padded sequences for subsequent text analysis. The model effectively combines both numerical and textual data, ensuring a holistic representation of products. A deep learning framework is constructed using TensorFlow and Keras, featuring multiple dense layers augmented with batch normalization and dropout techniques to enhance generalization and mitigate overfitting. Training is conducted using sparse categorical cross-entropy loss, with optimization performed via the Adam algorithm. The efficacy of the system is assessed through metrics such as accuracy, R² scores. The model achieves competitive accuracy, with further validation through loss minimization to gauge overall performance. This initiative underscores the capabilities of deep learning in the realm of e-commerce recommendations, facilitating users in making informed purchasing choices based on data. By incorporating both numerical and textual attributes, it enriches the personalized shopping experience, thereby fostering a more effective and competitive e-commerce environment. The suggested framework illustrates how deep learning techniques can optimize product selection, ultimately serving the interests of both consumers and online retailers by providing accurate and dependable recommendations within an expanding digital marketplace.
Keywords :
Product, E-commerce, Neural network, R2 Score, Recommendation, Product Matching.
