2013 No. 3
Display Method:
2013, (3): 1-14.
Abstract(
PDF 1639KB
Abstract:
Computational advertising is a kind of advertising mechanism which has the capability to find the most suitable ads for given users and web content, so as to advertises them accurately. Therefore, estimating click-through rate (CTR) precisely makes significant difference in the efficiency of advertising on the Internet. Ad click-through rate prediction is to estimate CTR with click log, which is influenced by the nature features of ad, the position, the page information, user properties, the reputation of advertisers and such other factors. This paper is aimed to illustrate useful CTR prediction models, including CTR models for ads of abundant history data, CTR models for rare ads or new ads and some optimization models. Finally, the implementation methods with real data set were demonstrated as examples.
Computational advertising is a kind of advertising mechanism which has the capability to find the most suitable ads for given users and web content, so as to advertises them accurately. Therefore, estimating click-through rate (CTR) precisely makes significant difference in the efficiency of advertising on the Internet. Ad click-through rate prediction is to estimate CTR with click log, which is influenced by the nature features of ad, the position, the page information, user properties, the reputation of advertisers and such other factors. This paper is aimed to illustrate useful CTR prediction models, including CTR models for ads of abundant history data, CTR models for rare ads or new ads and some optimization models. Finally, the implementation methods with real data set were demonstrated as examples.
2013, (3): 15-25.
Abstract:
CTR (Click-Through Rate) prediction can be used to improve users satisfaction with respect to the presented online advertisements (ads) and support effective advertising. CTR prediction is the basis for personalized recommendation of online ads. It is also necessary to re-commend ads and predict their CTRs for the users that have no historical click-through records. In this paper, we adopted BN (Bayesian network), an important probabilistic graphical model, as the framework for representing and inferring the similarity and the corresponding uncertainty of the behaviors in ad search of different users. First, we constructed the BN to reflect the similarity between users by means of statistic computations on the historical records of users ad search. Then, we measured the behavior similarity between the users with click-through records and those without records quantitatively based on the mechanism of BNs probabilistic inferences. Consequently, we predicted the CTRs of ads with respect to the users without historical click-through records, in order to provide a metric for ad recommendation. We made experiments on the training data of Tencent CA from KDD Cup 2012-Track 2 and tested the effectiveness of our methods.
CTR (Click-Through Rate) prediction can be used to improve users satisfaction with respect to the presented online advertisements (ads) and support effective advertising. CTR prediction is the basis for personalized recommendation of online ads. It is also necessary to re-commend ads and predict their CTRs for the users that have no historical click-through records. In this paper, we adopted BN (Bayesian network), an important probabilistic graphical model, as the framework for representing and inferring the similarity and the corresponding uncertainty of the behaviors in ad search of different users. First, we constructed the BN to reflect the similarity between users by means of statistic computations on the historical records of users ad search. Then, we measured the behavior similarity between the users with click-through records and those without records quantitatively based on the mechanism of BNs probabilistic inferences. Consequently, we predicted the CTRs of ads with respect to the users without historical click-through records, in order to provide a metric for ad recommendation. We made experiments on the training data of Tencent CA from KDD Cup 2012-Track 2 and tested the effectiveness of our methods.
2013, (3): 26-36，53.
Abstract:
The matching of the products features and users preferences is the key issue of the Internet Advertising. This paper introduced several important database query techniques bringing convenience for the information matching, provides their definitions and extended SQL, and the emphasis would be on reverse rank-aware queries. In the mean time, an example to demonstrate these query techniques was given. At last, the conclusion and vision were presented.
The matching of the products features and users preferences is the key issue of the Internet Advertising. This paper introduced several important database query techniques bringing convenience for the information matching, provides their definitions and extended SQL, and the emphasis would be on reverse rank-aware queries. In the mean time, an example to demonstrate these query techniques was given. At last, the conclusion and vision were presented.
2013, (3): 37-45.
Abstract:
Recommender systems aim to identify content of interest from overloaded information by exploiting the opinions of a community of users. Due to the complexity of spatial data and the unclear roles of context-aware information, developing personalized recommender systems in mobile and pervasive environments is more challenging than developing recommender systems from traditional domains. This paper introduced classic recommendation techniques and unique features in mobile recommender systems, as well as the challenges in mobile enviroment. Based on two cases, taxi driving route recommendation and personalized travel package recommendation, we formulated the mobile sequential recommendation (MSR) problem and constrained travel recommendation. Finally, we gave a brief solution of the mobile recommender problem respectively.
Recommender systems aim to identify content of interest from overloaded information by exploiting the opinions of a community of users. Due to the complexity of spatial data and the unclear roles of context-aware information, developing personalized recommender systems in mobile and pervasive environments is more challenging than developing recommender systems from traditional domains. This paper introduced classic recommendation techniques and unique features in mobile recommender systems, as well as the challenges in mobile enviroment. Based on two cases, taxi driving route recommendation and personalized travel package recommendation, we formulated the mobile sequential recommendation (MSR) problem and constrained travel recommendation. Finally, we gave a brief solution of the mobile recommender problem respectively.
2013, (3): 46-53.
Abstract:
For e-commerce sites, in order to promote the development and win more benefits, attracting and keeping the customers becomes very important. One of the most useful technologies is recommendation algorithms. In e-commerce sites, sidebar advertising is a common form of recommendation, which can be divided into three main categories: content-based, collaborative filtering and hybrid recommendation algorithms. However, current recommendation algorithms are challenged by new application requirements, such as diversification, personalization, intelligentization and timeliness. It is urgent to design new algorithms to meet these requirements.
For e-commerce sites, in order to promote the development and win more benefits, attracting and keeping the customers becomes very important. One of the most useful technologies is recommendation algorithms. In e-commerce sites, sidebar advertising is a common form of recommendation, which can be divided into three main categories: content-based, collaborative filtering and hybrid recommendation algorithms. However, current recommendation algorithms are challenged by new application requirements, such as diversification, personalization, intelligentization and timeliness. It is urgent to design new algorithms to meet these requirements.
2013, (3): 54-59，69.
Abstract:
With the rapid development of online advertising, the competition in Ad marketing is becoming more and more intense. It is important to accurately target for enhancing the ROI of online advertising. This paper proposes the definitions of the eigenvector space and eigenvector model of web pages for online advertising.Based on these concepts, it presents a fuzzy vector representation and its generating method for online advertising web pages. To accurately target advertising, it exploits a fuzzy clustering approach to partition web pages. A group of experiments has been conducted to verify the advertising efficiency and effectiveness with the proposed approach.
With the rapid development of online advertising, the competition in Ad marketing is becoming more and more intense. It is important to accurately target for enhancing the ROI of online advertising. This paper proposes the definitions of the eigenvector space and eigenvector model of web pages for online advertising.Based on these concepts, it presents a fuzzy vector representation and its generating method for online advertising web pages. To accurately target advertising, it exploits a fuzzy clustering approach to partition web pages. A group of experiments has been conducted to verify the advertising efficiency and effectiveness with the proposed approach.
2013, (3): 60-69.
Abstract:
How to get the maximum profit with online advertising is the most concern issue for advertisers. However, only pursuing the Click Through Rate maybe can not bring the maximum profit, which even induces huge amounts of network fraud information and fraudulent transaction. Considering the factors of the advertisers reputation, matching degree of product and clicking value, based on the mechanism design theory, an optimal launching mechanism for online advertising was proposed here, Further, the quality measure of advertise launching was calculated based on our mechanism. Through the final simulation testing and visual analysis, it was shown that our mechanism considering reputation has superiority on the launching efficiency.
How to get the maximum profit with online advertising is the most concern issue for advertisers. However, only pursuing the Click Through Rate maybe can not bring the maximum profit, which even induces huge amounts of network fraud information and fraudulent transaction. Considering the factors of the advertisers reputation, matching degree of product and clicking value, based on the mechanism design theory, an optimal launching mechanism for online advertising was proposed here, Further, the quality measure of advertise launching was calculated based on our mechanism. Through the final simulation testing and visual analysis, it was shown that our mechanism considering reputation has superiority on the launching efficiency.
2013, (3): 70-78.
Abstract:
Online advertising becomes increasingly important in the E-commerce age. Compared to traditional advertising media such as newspaper and TV channels, online advertising has the advantage of better precise targeting of the ads consumers due to the flexibility of digit content recommendation and personalization. The application of targeting techniques boosts the accurate delivery of ads to interested users. In this paper, we discuss the advance of advertising targeting both in research communities and industry.We show that targeted delivery of the ads by means of user segmentation can greatly increase the CTR of the advertising.
Online advertising becomes increasingly important in the E-commerce age. Compared to traditional advertising media such as newspaper and TV channels, online advertising has the advantage of better precise targeting of the ads consumers due to the flexibility of digit content recommendation and personalization. The application of targeting techniques boosts the accurate delivery of ads to interested users. In this paper, we discuss the advance of advertising targeting both in research communities and industry.We show that targeted delivery of the ads by means of user segmentation can greatly increase the CTR of the advertising.
2013, (3): 79-83，105.
Abstract:
In most E-commerce web sites, key word search is an important way for users to find commodities, and calculation of the term weights is a major step for search engine to process queries. In order to overcome the shortness of existing methods in calculation of term weights in query processing, this paper proposed a new method, which combines importance with relevance of terms when calculating term weights. It effectively improves the correctness of the determination of the term weights, and is practical and reliable.
In most E-commerce web sites, key word search is an important way for users to find commodities, and calculation of the term weights is a major step for search engine to process queries. In order to overcome the shortness of existing methods in calculation of term weights in query processing, this paper proposed a new method, which combines importance with relevance of terms when calculating term weights. It effectively improves the correctness of the determination of the term weights, and is practical and reliable.
2013, (3): 84-92，105.
Abstract:
After elaborating on the classification and characteristics of online advertising, this paper pointed out three shortcomings of image ads: fewer creative advertising,poor information transfer efficiency and low conversion. Then, in order to overcome these shortcomings, the concept of Motu (Magic Pictures) technology was proposed. Motu technology makes image ads no longer be confined to the traditional banners, windows and other forms, so as to take full advantage of the resources about the sites images instead.Each image will become the source of the advertising revenue to increase the effective impressions and improve advertising conversion efficiency.With the minimum advertising investment, outstanding user experience and significant advertisement effect,Motu technology fully meets the concept of images are advertising.
After elaborating on the classification and characteristics of online advertising, this paper pointed out three shortcomings of image ads: fewer creative advertising,poor information transfer efficiency and low conversion. Then, in order to overcome these shortcomings, the concept of Motu (Magic Pictures) technology was proposed. Motu technology makes image ads no longer be confined to the traditional banners, windows and other forms, so as to take full advantage of the resources about the sites images instead.Each image will become the source of the advertising revenue to increase the effective impressions and improve advertising conversion efficiency.With the minimum advertising investment, outstanding user experience and significant advertisement effect,Motu technology fully meets the concept of images are advertising.
2013, (3): 93-105.
Abstract:
Online advertising has sprung up and shown its diversity for decades. With the rapid increase of online advertising markets,a growing numbers of advertisers want to deliver ads related to their products or services to specific users with less expenditure. They expect these users have great interest in their products or services and have good intersection with ads(users see the ads, click it, registration or buy the products). Thus the concept of advertising target is proposed. This chapter discusses several classification of advertising target and for each category we elaborate on its application scenarios. We then summary and contrast the important advertising target methods and models used in recent years.
Online advertising has sprung up and shown its diversity for decades. With the rapid increase of online advertising markets,a growing numbers of advertisers want to deliver ads related to their products or services to specific users with less expenditure. They expect these users have great interest in their products or services and have good intersection with ads(users see the ads, click it, registration or buy the products). Thus the concept of advertising target is proposed. This chapter discusses several classification of advertising target and for each category we elaborate on its application scenarios. We then summary and contrast the important advertising target methods and models used in recent years.
2013, (3): 106-117.
Abstract:
This paper introduced the background and related techniques of online advertising market. While different types of online ads may imply different computational advertising techniques in the backend system, this paper provided a rational classification method of online advertising formats. The category of online advertising given in this paper covers existing and potential types in the domain. Further, the advertising platform have evolved in several major phases or generations, particularly ad server, ad network and ad exchange. We showed the system architectures of the advertising systems mentioned above, by discussing the main function modules and interfaces. Our work aims to give a comprehensive and detailed description of the online advertising systems from the view of computation. Besides, the surveys in our work can provide essential background knowledge for computational advertising related research.
This paper introduced the background and related techniques of online advertising market. While different types of online ads may imply different computational advertising techniques in the backend system, this paper provided a rational classification method of online advertising formats. The category of online advertising given in this paper covers existing and potential types in the domain. Further, the advertising platform have evolved in several major phases or generations, particularly ad server, ad network and ad exchange. We showed the system architectures of the advertising systems mentioned above, by discussing the main function modules and interfaces. Our work aims to give a comprehensive and detailed description of the online advertising systems from the view of computation. Besides, the surveys in our work can provide essential background knowledge for computational advertising related research.
2013, (3): 118-130.
Abstract:
In a careful study of Latent Dirichlet Allocation (LDA) using Gibbs sampling and the MapReduce framework, an efficient implementation for LDA in Mahout was achieved. The experiments showed the high performance of this distributed parallel LDA program, and several issues about enhancing performance were discussed.
In a careful study of Latent Dirichlet Allocation (LDA) using Gibbs sampling and the MapReduce framework, an efficient implementation for LDA in Mahout was achieved. The experiments showed the high performance of this distributed parallel LDA program, and several issues about enhancing performance were discussed.
2013, (3): 131-139，148.
Abstract:
Based on traditional homotopy analysis method (HAM), this method (PHAM) introduced a so-called prescribed parameter and associated condition to prove whether a nonlinear differential system admits multiple solutions. The PHAM was applied to two nonlinear differential systems and multiple solutions of associated system were obtained.
Based on traditional homotopy analysis method (HAM), this method (PHAM) introduced a so-called prescribed parameter and associated condition to prove whether a nonlinear differential system admits multiple solutions. The PHAM was applied to two nonlinear differential systems and multiple solutions of associated system were obtained.
2013, (3): 140-148.
Abstract:
Singularly perturbed systems for which the reduced system has a manifold of solutions are called singular singularly perturbed. Boundary value problems for such systems were examined by geometric singular perturbation approach in this paper. Assumptions were derived which ensure the existence of a locally unique solution which is near a singular orbit of the dynamics of limiting fast and slow systems.
Singularly perturbed systems for which the reduced system has a manifold of solutions are called singular singularly perturbed. Boundary value problems for such systems were examined by geometric singular perturbation approach in this paper. Assumptions were derived which ensure the existence of a locally unique solution which is near a singular orbit of the dynamics of limiting fast and slow systems.
2013, (3): 149-163，175.
Abstract:
This paper considered the global non-existence of solutions of nonlinear $p(x)$-Kirchhoff systems with dynamic boundary conditions, which involve nonlinear external damping terms $Q$ and nonlinear driving forces $f$. Through the study of the natural energy associated to the solutions $u$ of the systems, the nonexistence of global solutions, when the initial energy is controlled above by a critical value was proved. And the $p$-Kirchhoff equations involving the quasilinear homogeneous $p$-Laplace operator were extended to the $p(x)$-Kirchhoff equations which have been used in the last decades to model various phenomena.
This paper considered the global non-existence of solutions of nonlinear $p(x)$-Kirchhoff systems with dynamic boundary conditions, which involve nonlinear external damping terms $Q$ and nonlinear driving forces $f$. Through the study of the natural energy associated to the solutions $u$ of the systems, the nonexistence of global solutions, when the initial energy is controlled above by a critical value was proved. And the $p$-Kirchhoff equations involving the quasilinear homogeneous $p$-Laplace operator were extended to the $p(x)$-Kirchhoff equations which have been used in the last decades to model various phenomena.
2013, (3): 164-168，175.
Abstract:
The authors first obtained formulae for the determinants of reciprocal power GCD matrix and reciprocal power LCM matrix defined on a set $S$ consisting of three quasi-coprime divisor chains with $\gcd (S)\in S$. Consequently, it was showed that the reciprocal power GCD matrix and the reciprocal power LCM matrix defined on $S$ with $\gcd (S)\in S$ are nonsingular. But such result is not true when $S$ consists of three divisor chains which are not quasi-coprime.
The authors first obtained formulae for the determinants of reciprocal power GCD matrix and reciprocal power LCM matrix defined on a set $S$ consisting of three quasi-coprime divisor chains with $\gcd (S)\in S$. Consequently, it was showed that the reciprocal power GCD matrix and the reciprocal power LCM matrix defined on $S$ with $\gcd (S)\in S$ are nonsingular. But such result is not true when $S$ consists of three divisor chains which are not quasi-coprime.
2013, (3): 169-175.
Abstract:
The implicit function theorem under $t$-$P$-${\mathcal{K}}$-equivalence relation was obtained, and the $t$-$P$-${\mathcal{K}}$-codimension was also estimated. The result may be a powerful tool for the classification of smooth mapping germs with distinguished parameters relative to $t$-$P$-${\mathcal{K}}$-equivalence, and for the study of bifurcations of differential equation germs with complete integral.
The implicit function theorem under $t$-$P$-${\mathcal{K}}$-equivalence relation was obtained, and the $t$-$P$-${\mathcal{K}}$-codimension was also estimated. The result may be a powerful tool for the classification of smooth mapping germs with distinguished parameters relative to $t$-$P$-${\mathcal{K}}$-equivalence, and for the study of bifurcations of differential equation germs with complete integral.
2013, (3): 176-185，218.
Abstract:
In the process of optimization, the $Q$ factor of the converter, the $h$ ratio of magnetizing inductance to resonant inductance and the maximum flux $B_{\rm max}$ were chosen as the optimal parameters, the details on the power loss of the converter was presented, the boundary conditions was deduced and the process of getting optimum solution was proposed. Finally, a three-hundred-watt prototype was build; the theoretical and experimental comparison result verified the feasibility and accuracy of the design method.
In the process of optimization, the $Q$ factor of the converter, the $h$ ratio of magnetizing inductance to resonant inductance and the maximum flux $B_{\rm max}$ were chosen as the optimal parameters, the details on the power loss of the converter was presented, the boundary conditions was deduced and the process of getting optimum solution was proposed. Finally, a three-hundred-watt prototype was build; the theoretical and experimental comparison result verified the feasibility and accuracy of the design method.
2013, (3): 186-193.
Abstract:
In frame of the scattering-matrix theory and Fermi-Thomas approximation, we had investigated the conductance properties of mesoscopic structure in dc case by employing a quantum tunneling structure with single barrier. Results clearly indicate that, the contact effect has notable affection on the transmission probability of electrons and the distribution of internal potential in mesoscopic system; even though dc case, the behaviors of conductance displayed in potential drop conflict with the Kirchhoff's laws in classical circuit, and the total resistance of the mesoscopic system cannot be expressed as contact resistance in series with 'scatterer' resistance. So the quantum coherence between contacts and devices is worth considering, the role of contacts effect is important for further investigation of the electron transport in mesoscopic system.
In frame of the scattering-matrix theory and Fermi-Thomas approximation, we had investigated the conductance properties of mesoscopic structure in dc case by employing a quantum tunneling structure with single barrier. Results clearly indicate that, the contact effect has notable affection on the transmission probability of electrons and the distribution of internal potential in mesoscopic system; even though dc case, the behaviors of conductance displayed in potential drop conflict with the Kirchhoff's laws in classical circuit, and the total resistance of the mesoscopic system cannot be expressed as contact resistance in series with 'scatterer' resistance. So the quantum coherence between contacts and devices is worth considering, the role of contacts effect is important for further investigation of the electron transport in mesoscopic system.
2013, (3): 194-201.
Abstract:
By using method of wet chemical etching, a kind of straight aligned silicon nanowires (SiNWs) was prepared, with average length and diameter about 20 m and 100 nm, respectively. Then 1D SiNWs were used as the nano-electrode of capacitive ionization structure in a test system of ionization, to test the full volt ampere characteristics of ionization. The test results show that the as-grown SiNWs acting as one of the nano-electrode can reduce the breakdown voltage of power system greatly, because the as-grown SiNWs have large field enhancement factor, small size effect and high defect density.
By using method of wet chemical etching, a kind of straight aligned silicon nanowires (SiNWs) was prepared, with average length and diameter about 20 m and 100 nm, respectively. Then 1D SiNWs were used as the nano-electrode of capacitive ionization structure in a test system of ionization, to test the full volt ampere characteristics of ionization. The test results show that the as-grown SiNWs acting as one of the nano-electrode can reduce the breakdown voltage of power system greatly, because the as-grown SiNWs have large field enhancement factor, small size effect and high defect density.
2013, (3): 202-208.
Abstract:
A nonenzymatic glucose sensor was fabricated by electroplating palladium particles on a nickel modified silicon nanowires(Ni/SiNWs) array electrode. The morphology of the Pd-Ni/SiNWs electrode was characterized by scanning electron microscope (SEM) and energy-dispersive X-ray spectroscopy (EDS) analysis. The performance of this sensor was characterized by cyclic voltammetry (CV) and fixed potential amperometry techniques. Under the fixed potential of +0.19 V, the sensitivity of the electrochemical oxidation for glucose is 302.2 AmM-1cm-2 and the detection limit (S/N ratio=3) is 4.9 M in the 0.1 M KOH alkaline electrolyte. Furthermore, the electrode shows good resistance for the interfering species such as the ascorbic acid (AA) and uric acid (UA). This novel sensor has great potential for the determination of glucose.
A nonenzymatic glucose sensor was fabricated by electroplating palladium particles on a nickel modified silicon nanowires(Ni/SiNWs) array electrode. The morphology of the Pd-Ni/SiNWs electrode was characterized by scanning electron microscope (SEM) and energy-dispersive X-ray spectroscopy (EDS) analysis. The performance of this sensor was characterized by cyclic voltammetry (CV) and fixed potential amperometry techniques. Under the fixed potential of +0.19 V, the sensitivity of the electrochemical oxidation for glucose is 302.2 AmM-1cm-2 and the detection limit (S/N ratio=3) is 4.9 M in the 0.1 M KOH alkaline electrolyte. Furthermore, the electrode shows good resistance for the interfering species such as the ascorbic acid (AA) and uric acid (UA). This novel sensor has great potential for the determination of glucose.
2013, (3): 209-218.
Abstract:
This paper took an improved algorithm based on inter-class separability directed acyclic graph support vector machine (DAG-SVM) for text classification.The method has adjusted the DAG structure according to inter-class distribution and the distance between centers. It has solved the problems of fixed structure and random single node location in traditional DAG-SVM multi-classification method.The experiments show that the algorithm has improved the accuracy.
This paper took an improved algorithm based on inter-class separability directed acyclic graph support vector machine (DAG-SVM) for text classification.The method has adjusted the DAG structure according to inter-class distribution and the distance between centers. It has solved the problems of fixed structure and random single node location in traditional DAG-SVM multi-classification method.The experiments show that the algorithm has improved the accuracy.