In recent years, with the rapid development of mobile Internet and its business applications, mobile advertising Click-Through Rate (CTR) estimation has become a hot research direction in the field of computational advertising, which is used to achieve accurate advertisement delivery for the best benefits in the three-side game between media, advertisers, and audiences. Current research on the estimation of CTR mainly uses the methods and models of machine learning, such as linear model or recommendation algorithms. However, most of these methods are insufficient to extract the data features and cannot reflect the nonlinear relationship between different features. In order to solve these problems, we propose a new model based on Deep Belief Nets to predict the CTR of mobile advertising, which combines together the powerful data representation and feature extraction capability of Deep Belief Nets, with the advantage of simplicity of traditional Logistic Regression models. Based on the training dataset with the information of over 40 million mobile advertisements during a period of 10 days, our experiments show that our new model has better estimation accuracy than the classic Logistic Regression (LR) model by 5.57% and Support Vector Regression (SVR) model by 5.80%.
Distributed Denial of Service (DDoS) attack presents the following characteristics, that the botnets become extra-large scale, the mode of attack presents a variety of characteristics and the application-level attacks become the main attack approach, which seriously impact on Internet Security. However, traditional software defense detection means have such problem, that the accurate rate is too low, detecting method is excessively obsolete and detecting way is excessively passive and the deployment of defense system is cumbersome. While hardware defense system such as ACL and IDMS products costs much, which small or medium-sized website has no ability to bear it. For the above reasons, we try to use artificial intelligence methods. Using the Turing test method to detect users, who do the behavior. Using modified RBF neural network to detect attack, designing intelligent user control system to deal with the complex and ever-changing attacks. The test results show that this defense system cost lowly, own strong defense capability, has the ability to deal with the current distributed denial of service attacks and impact on the server running performance less.
Currently, collaborative filtering (CF) is one of the most popular and successful recommendation technologies. However, the algorithm is easily affected by data sparsity, leading to poor recommendation accuracy. Recently, Deep Belief Nets (DBNs) have been successfully applied in many research areas including image classification and phone recognition. In this paper, to solve the data sparsity problem in CF, we propose a hybrid recommendation model based on DBNs and K-nearest neighbor (KNN) algorithm in which the user-based KNN algorithm makes predictions using the user features extracted by the DBNs. We also present efficient learning and inference methods for this hybrid model and demonstrate that DBNs can be successfully applied to CF. Finally, we carried out several experiments on MovieLens dataset which demonstrate that our hybrid model can achieve better recommendation results than some other CF methods which are widely used.
After decryption, the data in mobile storage is lack of the protection from encryption. To solve this problem, our team proposed an aprotection scheme for classified documents, which is based on kernel level file filtering system, implements a set of data protection system for mobile storage medium. First, through the transparent encryption technology, ensure that unauthorized visitors cannot access classified documents; Then through the file redirection technology, make sure that the decrypted data is under the comprehensive protection when authorized users are in the normal using process; Finally, clear the temporary file on the disk and the data trace in the memory, ensure that no using trace left on the computer. Experimental results show that this project has realized comprehensive security protection for classified documents, and achieve the goal that classified devices can be used on unclassified computer safely.
In the Middle-size League of RoboCup, the situation on the ground is unpredictable as the football moves irregularly whether viewed from the speed or the direction or the height of the football movement. Therefore, a single defensive strategy cannot meet the demands of the goalkeeper for the goalkeeping. In view of this, this paper proposes the use of dynamic vision and 3D imaging technology, which solves the problem of insufficient capacity of the goalkeeper using a single strategy according to the football area and height and different strategic options, thus greatly improving the success rate of the gatekeepers. In addition, the feasibility of the strategy is also verified through the actual match.