H.3. Artificial Intelligence
Morteza Mohammadi Zanjireh; Nasrin Nazari
Abstract
Predicting customer purchase intention is an important yet challenging task in e-commerce due to the sequential nature of user interactions and class imbalance in real-world shopping data. This study presents a purchase prediction framework that combines conventional session-based behavioral features ...
Read More
Predicting customer purchase intention is an important yet challenging task in e-commerce due to the sequential nature of user interactions and class imbalance in real-world shopping data. This study presents a purchase prediction framework that combines conventional session-based behavioral features with semantic similarity measures derived from learned product embeddings. Experiments were conducted on a large-scale online cosmetics store dataset containing more than 12 million user interaction events. After preprocessing and session construction, product sequences were represented using four embedding techniques: Word2Vec, Doc2Vec, GloVe, and FastText. Rather than directly using embedding vectors as predictive inputs, the proposed approach derives semantic similarity measures from product representations and integrates them with conventional behavioral attributes. To address class imbalance, SMOTE and RandomUnderSampler were evaluated. The resulting feature sets were assessed using six machine learning classifiers: Decision Tree, Random Forest, Bagging, AdaBoost, Logistic Regression, and XGBoost. The results show that semantic similarity features generally improve prediction performance across the evaluated configurations. Doc2Vec achieved the strongest overall performance among the investigated embedding techniques, while Random Forest and Bagging performed particularly well. Additional ROC-AUC and PR-AUC analyses further support the usefulness of the proposed similarity-based representation beyond the primary F1-score evaluation. Overall, the findings indicate that semantic relationships among products provide complementary information to conventional behavioral features and can contribute to improved purchase prediction in e-commerce environments.
H.5. Image Processing and Computer Vision
Reza Kharghanian; Zeynab Mohammadpoory
Abstract
The Convolutional Restricted Boltzmann Machine (CRBM) is a generative model that extracts representations from unlabeled data, achieving success in various applications. However, its unsupervised nature may yield suboptimal representations for specific classification tasks. This paper proposes adapting ...
Read More
The Convolutional Restricted Boltzmann Machine (CRBM) is a generative model that extracts representations from unlabeled data, achieving success in various applications. However, its unsupervised nature may yield suboptimal representations for specific classification tasks. This paper proposes adapting k-means clustering to enhance CRBM parameters, aligning features with informative cluster centers. A novel criterion combining generative and soft-K-Means objectives optimizes both cluster centers and CRBM parameters, allowing for continued unsupervised feature learning.Experiments on MNIST, CIFAR10, and three facial expression datasets (JAFFE, KANADE, BU) show that the proposed method enhances the learning process and offers a more informative representation compared to standard and classification CRBM.
