Session interest model for CTR prediction based on self-attention mechanism

Click-through rate prediction, which aims to predict the probability of the user clicking on an item, is critical to online advertising. How to capture the user evolving interests from the user behavior sequence is an important issue in CTR prediction. However, most existing models ignore the factor that the sequence is composed of sessions, and user behavior can be divided into different sessions according to the occurring time. The user behaviors are highly correlated in each session and are not relevant across sessions. We propose an effective model for CTR prediction, named Session Interest Model via Self-Attention (SISA). First, we divide the user sequential behavior into session layer. A self-attention mechanism with bias coding is used to model each session. Since different session interest may be related to each other or follow a sequential pattern, next, we utilize gated recurrent unit (GRU) to capture the interaction and evolution of user different historical session interests in session interest extractor module. Then, we use the local activation and GRU to aggregate their target ad to form the final representation of the behavior sequence in session interest interacting module. Experimental results show that the SISA model performs better than other models.

Research on News Recommendation System Based on Deep Network and Personalized Needs

In order to solve the problems of poor performance of the recommendation system caused by not considering the needs of users in the process of news recommendation, a news recommendation system based on deep network and personalized needs is proposed. Firstly, it analyzes the news needs of users, which is the basis of designing the system. The functions of the system module mainly include the network function module, database module, user management module, and news recommendation module. Among them, the user management module uses the deep network to set the user news interest model, inputs the news data into the model, completes the personalized needs of the news, and realizes the design of the news recommendation system. The experimental results show that the proposed system has good effect and certain advantages.

“Was that Interesting?” & “Does it Matter?”: The implications of On-Task learning experiences

Interest is a critical fuel for and outcome of learning. Building on and refocusing the Four- Phase Model of interest development, this study provides a window into the ecology of the learning experience and interest it generates. This research tests a task interest model for understanding learning experiences in three university courses (mathematics, biochemistry and introduction to physics for non-majors) and pilots a micro-analytic approach to capturing these experiences during lectures/tutorials. Students' interest in tasks, a single class and the domain of study were collected. Short surveys were conducted through an online platform during class, immediately following task experiences. Latent variance-based modelling suggested strong forward connections between interest in most tasks were observed. The connections between prior knowledge and interest with future interest in course tasks varied strongly and was dependent on the nature of the tasks. The nuance of these connections and their implications for theory and practice are discussed.

Research on the construction of O2O teaching system of cross-cultural knowledge in College English based on MOOC

In order to improve the teaching effect of college English cross-cultural knowledge, with the support of recommendation algorithms, this paper builds a O2O teaching system of cross-cultural knowledge in college English based on MOOC, and provides service methods such as material resource learning, “smart grab-order” question answering, resource recommendation, LBS appointment teaching and offline teaching. Moreover, this paper constructs a learner and teacher interest model, introduces a suspicious degree update model into the interest model, uses a deep learning network algorithm to recommend composition resources, and uses a deep network adaptive learning algorithm to enhance the accuracy of composition resource recommendation. In addition, this paper constructs a system framework structure based on actual needs, and studies the realization process of each of its functional modules. Finally, this paper designs an experiment to evaluate the practical effect of O2O teaching system of cross-cultural knowledge in college English based on MOOC. The research results show that the system constructed in this paper meets the expected requirements.

Personalized Mobile Information Recommendation Based on Fine-Grained User Behaviors

At present, the enthusiasm of users to score actively in mobile information recommendation system is generally poor. Moreover, the existing research works rarely start with the analysis of fine-grained reading behaviors of mobile terminal users, but mostly based on the analysis of reading content and the improvement of model. It is difficult to find out the objective, short-term and local behavioral preferences of users. To solve the above problems, we propose six kinds of explicit fine-grained reading behaviors and integrate them into the user reading interest model to form the SVR-ALL model. The effectiveness of these six fine-grained behaviors is verified by ablative experiments. On the basis of SVR-ALL model, four implicit fine-grained reading behaviors are further mined by considering the difference of user reading habits, and then propose the user reading preference model called F-AFC. The updating mechanism for user preference designed in F-AFC can fully reflect the changes of users’ reading habits in different periods. Experiments show that the accuracy of the user interest model considering user’s reading preference and its update can be improved to some extent.

DIM: Adaptively Combining User Interests Mined at Different Stages Based on Deformable Interest Model

User interest mining is widely used in the fields of personalized search and personalized recommendation. Traditional methods ignore the formation of user interest which is a process that evolves over time. This leads to the inability to accurately describe the distribution of user interest. In this paper, we propose the interest tracking model (ITM). To add the timing, ITM uses Dirichlet distribution and multinomial distribution to describe the evolutional process of interest topics and frequent patterns, which well adapts to the evolution of user interest hidden in short texts between different time slices. In addition, it is well known that user interest is composed of long-term interest and situational interest including short-term interest and social hot topics. State-of-the-art methods simply regard the users’ long-term interest as the users’ final interest, which makes those unable to completely describe the user interest distribution. To solve this problem, we propose the deformable interest model (DIM) which designs an objective function to combine users’ long-term interest and situational interest and more comprehensively and accurately mine user interest. Furthermore, we present the degree of deformation which measures the subinterest's degree of influence on final interest and propose in DIM the influence real-time update mechanism. The mechanism adaptively updates the degree of deformation through the linear iteration and reduces the degree of dependence of the interest model on training sets. We present results via a dataset consisting of Flickr users and their uploaded information in three months, a dataset consisting of Twitter users and their tweets in three months, and a dataset consisting of Instagram users and their uploaded information in three months, showing that the perplexity is reduced to 0.378, the average accuracy is increased to 94%, and the average NMI is increased to 0.20, which prove better interest prediction.