Junhao Wen

Dropout prediction in MOOCs using behavior features and multi-view semi-supervised learning.

While massive open online courses (MOOCs) have gained increasing popularity in recent years, dropout prediction has been an important task to solve due to the high rates of dropout students found in MOOCs. Current methods normally apply supervised learning methods to dropout prediction, using general features extracted from behavior records without the concern of behavior types. However, student learning behavior is diverse and there are no sufficient labeled data to train a model because it is a time-costing task to label enormous data in practice. To solve these problems, in this work, we proposed a novel multi-view semi-supervised learning model based on behavior features for the dropout prediction task. Specifically, we derived features from each type of learning behavior to form multi-view behavior features. In addition, based on these features, we proposed a new multi-view semi-supervised learning method to make use of a large number of unlabeled data to assist insufficient labeled data for improving prediction performance. We conducted experiments on KDD Cup 2015 dataset, and the results show that our proposed method achieves better prediction of student dropouts as compared to state-of-the-art approaches.

A Hybrid Trust Degree Model in Social Network for Recommender System

Recommender system is an effective way to help users to find the required information. In the social network, the recommendation is often from one user to another user. Therefore, it is necessary to determine how the two users trust each other. However, much work has paid more attention to the one-to-one trust relationship but ignored the many-to-one relationship. In this paper, we proposed a hybrid trust degree model to describe how two users trust each other. This model not only considers the direct trust degree and indirect trust degree between the two users, but also considers the group trust degree. The group trust degree describes how a user are trusted by other users in a group. The experiment result shows that hybrid trust degree can reasonably measure and calculate the credit between two users in a group.

LSSL-SSD: Social Spammer Detection with Laplacian Score and Semi-supervised Learning

The rapid development of social networks makes it easy for people to communicate online. However, social networks usually suffer from social spammers due to their openness. Spammers deliver information for economic purposes, and they pose threats to the security of social networks. To maintain the long-term running of online social networks, many detection methods are proposed. But current methods normally use high dimension features with supervised learning algorithms to find spammers, resulting in low detection performance. To solve this problem, in this paper, we first apply the Laplacian score method, which is an unsupervised feature selection method, to obtain useful features. Based on the selected features, the semi-supervised ensemble learning is then used to train the detection model. Experimental results on the Twitter dataset show the efficiency of our approach after feature selection. Moreover, the proposed method remains high detection performance in the face of limited labeled data.

Semi-supervised learning combining transductive support vector machine with active learning

In typical data mining applications, labeling the large amounts of data is difficult, expensive, and time consuming, if annotated manually. To avoid manual labeling, semi-supervised learning uses unlabeled data along with the labeled data in the training process. Transductive support vector machine (TSVM) is one such semi-supervised, which has been found effective in enhancing the classification performance. However there are some deficiencies in TSVM, such as presetting number of the positive class samples, frequently exchange of class label, and its requirement for larger amount of unlabeled data. To tackle these deficiencies, in this paper, we propose a new semi-supervised learning algorithm based on active learning combined with TSVM. The algorithm applies active learning to select the most informative instances based on the version space minimum-maximum division principle with human annotation for improve the classification performance. Simultaneously, in order to make full use of the distribution characteristics of unlabeled data, we added a manifold regularization term to the objective function. Experiments performed on several UCI datasets and a real world book review case study demonstrate that our proposed method achieves significant improvement over other benchmark methods yet consuming less amount of human effort, which is very important while labeling data manually.

Sales Growth Rate Forecasting Using Improved PSO and SVM

Accurate forecast of the sales growth rate plays a decisive role in determining the amount of advertising investment. In this study, we present a preclassification and later regression based method optimized by improved particle swarm optimization (IPSO) for sales growth rate forecasting. We use support vector machine (SVM) as a classification model. The nonlinear relationship in sales growth rate forecasting is efficiently represented by SVM, while IPSO is optimizing the training parameters of SVM. IPSO addresses issues of traditional PSO, such as relapsing into local optimum, slow convergence speed, and low convergence precision in the later evolution. We performed two experiments; firstly, three classic benchmark functions are used to verify the validity of the IPSO algorithm against PSO. Having shown IPSO outperform PSO in convergence speed, precision, and escaping local optima, in our second experiment, we apply IPSO to the proposed model. The sales growth rate forecasting cases are used to testify the forecasting performance of proposed model. According to the requirements and industry knowledge, the sample data was first classified to obtain types of the test samples. Next, the values of the test samples were forecast using the SVM regression algorithm. The experimental results demonstrate that the proposed model has good forecasting performance.

Item Anomaly Detection Based on Dynamic Partition for Time Series in Recommender Systems

In recent years, recommender systems have become an effective method to process information overload. However, recommendation technology still suffers from many problems. One of the problems is shilling attacks-attackers inject spam user profiles to disturb the list of recommendation items. There are two characteristics of all types of shilling attacks: 1) Item abnormality: The rating of target items is always maximum or minimum; and 2) Attack promptness: It takes only a very short period time to inject attack profiles. Some papers have proposed item anomaly detection methods based on these two characteristics, but their detection rate, false alarm rate, and universality need to be further improved. To solve these problems, this paper proposes an item anomaly detection method based on dynamic partitioning for time series. This method first dynamically partitions item-rating time series based on important points. Then, we use chi square distribution (χ2) to detect abnormal intervals. The experimental results on MovieLens 100K and 1M indicate that this approach has a high detection rate and a low false alarm rate and is stable toward different attack models and filler sizes.

Robust Social Recommendation Techniques: A Review

Social recommendation plays an important role in solving the cold start problem in recommendation systems and improves the accuracy of recommendation, but still faces serious challenges and problems. Ratings or relationships injected by fake users seriously affect the authenticity of the recommendations as well as users’ trustiness on the recommendation systems. Moreover, the simplification of relationship treatment also seriously affects the recommendation accuracy and user satisfaction to the recommendation systems. This paper first analyzes up to date research of social recommendation and the detecting technology of multiple relationships. Furthermore, it proposes a future research framework for robust social recommendations including modeling and feature extraction of multidimensional relationships, social recommendation shilling attack models based on social relationships, the analysis of the relationships in social networks as well as the roles of relationships on recommendation, and robust social recommendation approaches taking multiple relationships into consideration.

A Restaurant Recommender System Based on User Preference and Location in Mobile Environment

Recommender system is an effective way to help users to obtain the personalized and useful information. However, due to complexity and dynamic, the traditional recommender system cannot work well in mobile environment. In this paper, we propose a restaurant recommender system in mobile environment. This recommender system adopts a user preference model by using the features of users visited restaurants, and also utilizes the location information of user and restaurants to dynamically generate the recommendation results. Baidu map cloud service is used to implement the proposed recommender system. The result of a case study shows that the proposed restaurant recommender system can effectively utilize users preference and the location information to recommend the personalized and suitable restaurants for different users.

Social recommendation using Euclidean embedding

Traditional recommender systems assume that all the users are independent, and they usually face the cold start and data sparse problems. To alleviate these problems, social recommender systems use social relations as an additional input to improve recommendation accuracy. Social recommendation follows the intuition that people with social relationships share some kinds of preference towards items. Current social recommendation methods commonly apply the Matrix Factorization (MF) model to incorporate social information into the recommendation process. As an alternative model to MF, we propose a novel social recommendation approach based on Euclidean Embedding (SREE) in this paper. The idea is to embed users and items in a unified Euclidean space, where users are close to both their desired items and social friends. Experimental results conducted on two real-world data sets illustrate that our proposed approach outperforms the state-of-the-art methods in terms of recommendation accuracy.

Semi-supervised hybrid clustering by integrating Gaussian mixture model and distance metric learning

Semi-supervised clustering aim to aid and bias the unsupervised clustering by employing a small amount of supervised information. The supervised information is generally given as pairwise constraints, which was used to either modify the objective function or to learn the distance measure. Many previous work have shown that the cluster algorithm based on distance metric is significantly better than the cluster algorithm based on probability distribution in the some data set, there are a totally opposite result in another data set, so how to balance the two methods become a key problem. In this paper, we proposed a semi-supervised hybrid clustering algorithm that provides a principled framework integrating distance metric into Gaussian mixture model, which consider not only the intrinsic geometry information but also the probability distribution information of the data. In comparison to only using the pairwise constraints, the labeled data was used to initialize Gaussian distribution parameter and to construct the weight matrix of regularizer, and then we adopt Kullback-Leibler Divergence as the “distance” measurement to regularize the objective function. Experiments on several UCI data sets and the real world data sets of Chinese Word Sense Induction demonstrate the effectiveness of our semi-supervised cluster algorithm.