Search

With the rise of online shopping becoming an integral part of daily life, it has brought unparalleled convenience, allowing us to purchase anything from daily necessities to luxury items with ease. On platforms like Amazon[1], customer feedback mechanisms play a crucial role in shaping user behavior and business practices. These mechanisms include the Star Rate (1-5) and detailed reviews. The star rating provides a quick and intuitive way for customers to score a product, while reviews offer comprehensive descriptions and shopping experiences. These feedback systems influence other customers' purchasing decisions and provide valuable insights for businesses to improve their products. In our study, we explore the impact of time on user behavior by combining three datasets for analysis. We observe a macro trend of increasing online shopping activity and customer satisfaction over time, with a growing tendency for customers to leave feedback. From a micro perspective, we conduct time series analysis and establish that customer star ratings vary over time, forming a column relationship, which we model using the ARIMA[2] technique to predict future trends. Furthermore, we investigate the influence of reviews on customer responses, finding that negative reviews spread rapidly and widely through social networks, akin to a viral phenomenon. Utilizing the SIR virus model as a text-based network propagation model, we demonstrate the significant impact of negative reviews on consumption patterns.[3] Lastly, we employ advanced deep learning models such as BERT, LLM, and GPT for natural language processing (NLP) to analyze the sentiment in customer reviews. By leveraging word vectors through classification functions, we distinguish between pessimistic and optimistic emotions. Our findings reveal a higher-order functional relationship[4] between star ratings and these emotions, providing deeper insights into customer sentiment and product perception.

This paper discusses the importance of spam classification in the field of information security. With the popularity of the Internet and email, spam has become one of the major issues affecting user experience and information security. The study begins with preprocessing text data in various ways, including converting to lowercase, removing irrelevant content, links, punctuation, etc., and filtering deactivated words and words of length 1. By applying the DistilBERT model to the text classification task, the results show that it achieves 93% accuracy in spam classification, effectively distinguishing between spam and non-spam emails. The confusion matrix showed that 18,500 emails were correctly classified and a small number of spam emails were misclassified as non-spam emails. Overall, the DistilBERT model showed high accuracy in spam classification, but more algorithms are still expected to emerge to improve the prediction accuracy. This study provides a useful reference for improving spam filtering systems in the future, which is expected to further enhance user experience and information security.

Our work presents a neural network designed to convert textual descriptions into 3D models. By leveraging the encoder-decoder architecture, we effectively combine text information with attributes such as shape, color, and position. This combined information is then input into a generator to predict new furniture objects, which are enriched with detailed information like color and shape.[1] The predicted furniture objects are subsequently processed by an encoder to extract feature information, which is then utilized in the loss function to propagate errors and update model weights. After training the network, we can generate new 3D objects solely based on textual input, showcasing the potential of our approach in generating customizable 3D models from descriptive text.[2]

In this project, we present a novel approach to depth perception using a monocular camera by incorporating information from both RGB and LiDAR modalities. Our primary objective is to investigate the performance and effectiveness of different techniques to generate accurate depth estimation. We first implemented the Swin Transformer-based depth estimation model and evaluated its performance on KITTI dataset containing RGB images and their corresponding ground truth depth maps. Next, we proposed an RGB-LiDAR fusion model. We performed necessary preprocessing steps on the dataset, such as resizing, normalization, and data augmentation, and trained both models with identical configurations for a fair comparison. Our results demonstrate that the proposed RGB- LiDAR fusion model achieves superior depth estimation performance compared to the original Swin Transformer based model. We evaluated the models on the test dataset using metrics such as mean absolute error (MAE) and root mean squared error (RMSE). The enhanced performance indicates the potential benefits of RGB-LiDAR fusion for monocular depth perception tasks. This study offers valuable insights into the strengths [1] and weaknesses of combining RGB and LiDAR inputs and lays the foundation for future research in monocular depth perception, aiming to further improve model architectures and training techniques.