Zhichao Yin

A New SVM Method for Short Text Classification Based on Semi-Supervised Learning

Short text is a popular text form, which is widely used in short commentary, micro-blog and many other fields. With the development of the social software and movie websites, the size of data is also becoming larger and larger. Most data is useless for us while other data is important for us. Therefore, it is very necessary for us to extract the useful short text from the big data. However, there are some problems such as fewer features, irregularity on the short text classification. To solve the problem we should pretreat the short text set and choose the significant features. This paper use semi-supervised learning and SVM to improve the traditional method and it can classify a large number of short texts to mining the useful massage from the short text. The experimental results also show a good improvement.

Anomaly detection model based on data stream clustering

Intrusion detection provides important protection for network security and anomaly detection as a type of intrusion detection, which can recognize the pattern of normal behaviors and label the behaviors which departure from normal pattern as anomaly behaviors. The updating of network equipment and broadband speed makes the data mining object change from static data sets to dynamic data streams. We think that the traditional methods based on data set do not satisfy the needs of dynamic network environment. The network data stream is temporal and cannot be treated as static data set. The concept and distribution of data objects is variety in different time stamps and the changing is unpredictable. Therefore, we propose an improved data stream clustering algorithm and design the anomaly detection model according to the improved algorithm. The established model can be modified with the changing of data stream and detect anomaly behaviors in time.

A Hybrid Feature Selection Algorithm

Feature selection algorithm in intrusion detection, data mining and pattern recognition plays a crucial role, it deletes unrelated and redundant features of the original data set to the optimal feature subset which are applied to some evaluation criteria. Due to the low accuracy, the high false positive rate and the long detection time of the existing feature selection algorithm, in the paper, we put forward a hybrid feature selection algorithm towards efficient intrusion detection, this algorithm chooses the optimal feature subset by combining the correlation algorithm and redundancy algorithm. Experimental results show that the algorithm shows almost and even better than the traditional feature selection algorithm on the different classifiers.

A Novel Framework towards Botnet Detection

Botnet have turned into the most serious security dangers on the present Internet framework. In this paper proposed an anomaly detection model based on genetic neural network system, which joined the significant global searching capability of genetic algorithm with the precise local searching element of back propagation, feed forward neural networks to improve the initial weights of neural network.

An Improved Trust Model Based on Time Effect

Trust model is a very important model in social networking and recommendation technology. The trust relationship between the users can reflect the relation in the real life in a better performance than the similarity, but this kind of trust model lose sight of the importance of time effect, so this article pay attention to the research on time effect in trust model and proposed an improved trust model based on time effect. At last we choose appropriate data set to prove the superiority of the proposed improved trust model.

Text Classification Algorithm Based on SLAS-C

Nowadays, mobile marketing is becoming increasingly important both strategically and economically because of the mobile devices. Short text is becoming a popular text form which can be seen in many fields such as network news, QQ messages, comments in BBS and so forth. Besides, our mobile devices also contain a lot of data of short text. To extract useful information from the short text more efficiently, this paper proposes SLAS (semi-supervised learning method and SVM classifier) and CART (classification and regression tree) to improve the traditional methods, which can classify massive short texts to mining the useful information from the short texts. The experiment also shows a better result than before, which has a more than 10% increase, including precision rate, recall rate and F1 value, besides, the running time is reduced by half than the KNN algorithm.

A Feature Selection Algorithm of Dynamic Data-Stream Based on Hoeffding Inequality

With the rapid development of the Internet, the application of data mining in the Internet is becoming more and more extensive. However, the complex data sources features are making the data mining process become very inefficient. In order to make data mining more efficient and simple, feature selection research is essential. In this paper, a new metric of mutual information based on mutual information is proposed (measure the correlation degree of the internal features of the collection), additionally Hoeffding inequality is also introduced to construct the HSF algorithm. The HSF is compared with the BIF (based on mutual information feature selection algorithm), the C4.5 classification algorithm is used as the testing algorithm for the experiments. Experiments show that HSF has better performance than BIF [1] in classification accuracy and error rate.

A New Mobile Recommendation Algorithm Based on Statistical Theory

With Recommendation technology has been widely used in advertising push, e-commerce and other fields and it has shown its powerful application prospect. But with the index increasing of mobile commerce data size, the size of the recommendation system is also increased and this leads to that the traditional collaborative filtering recommendation algorithm cannot adapt to such a big data processing. To solve the problem, we proposed an algorithm based on the statistical analysis of user data. First, this algorithm classified the data simply, and then we could gain the relatively accurate personalized recommendation results by the statistical analysis of different attributes on the data sets.