E-commerce recommender systems (RSs) can help users quickly find what they need or new products they might be interested in. To continuously enhance user trust in the website, improve page visits and dwell time, and most importantly, increase gross merchandise value (GMV), it is crucial to understand and capture the important information hidden in the data, which has a great impact on user choice. The fashion e-commerce websites can collect the attributes of items and users as well as the user purchase behaviors, but lack the fine-grained classification of the items and the implicit relationship between items and users. This paper focuses on Amazon fashion dataset, one of the most widely used datasets in the fashion field. A differentiated recommendation framework is proposed that provides different recommendation paths for active and inactive users to improve the overall recommendation quality. In the framework, a data augmentation algorithm based on transfer learning is proposed to filter out the irrelevant items and label items with fine-grained tags, a user-item knowledge graph is built to discover the potential relationship between items and users. Finally, a differentiated recommendation strategy is put forward to make different recommendations for users with different characteristics. The experimental results show that through data augmentation algorithm to improve data quality, factorization machine model produces higher recommendation accuracy, the constructed knowledge graph can alleviate the cold start problem for recommendation, and the differentiated recommendation strategy has achieved better recommendations for both active and inactive users.

Clothing images vary in style and everyone has a different understanding of style. Even with the current popular deep learning methods it is difficult to accurately classify style labels. A style representation learning model based on the deep neural networks called StyleNet is proposed in this paper. We adopt multi-task learning framework to build the model and make full use of various types of label information to represent the clothing images in a finer-grained manner. Due to the semantic abstraction of image labels in the current fashion field, using a simple migration learning method cannot fully meet the requirements of clothing image classification. An objective function optimization method is put forward by combining the distance confusion loss and the traditional cross entropy loss to improve the accuracy of StyleNet further. The experimental results show that by applying the multi-task representation learning framework StyleNet can achieve better classification accuracy, the optimized loss function can also bring performance improvement for deep learning models, and the classification effect of StyleNet becomes better as the size of the data set increases. In order to verify the robustness and effectiveness of the deep learning method in StyleNet, we also apply Faster R-CNN module to pre-process the clothing images and use the result as the input of StyleNet. The classifier can only get a limited performance improvement, which is negligible compared to the methods proposed in this paper of increasing the depth of the neural network and optimizing the loss function.