Search results

Analysis of the publisher's behavior plays a vital role in identifying fraudulent publishers in the pay-per-click model of online advertising. However, the vast amount of raw user click data with missing values pose a challenge in analyzing the conduct of publishers. The presence of high cardinality in categorical attributes with multiple possible values has further aggrieved the issue.

In this paper, gradient tree boosting (GTB) learning is used to address the challenges encountered in learning the publishers' behavior from raw user click data and effectively classifying fraudulent publishers.

The results demonstrate that the GTB effectively classified fraudulent publishers and exhibited significantly improved performance as compared to other learning methods in terms of average precision (60.5 %), recall (57.8 %) and f-measure (59.1%).

The experiments were conducted using publicly available multiclass raw user click dataset and eight other imbalanced datasets to test the GTB's generalizing behavior, while training and testing were done using 10-fold cross-validation. The performance of GTB was evaluated using average precision, recall and f-measure. The performance of GTB learning was also compared with eleven other state-of-the-art individual and ensemble classification models.

The problem of choosing the utmost useful features from hundreds of features from time-series user click data arises in online advertising toward fraudulent publisher's classification. Selecting feature subsets is a key issue in such classification tasks. Practically, the use of filter approaches is common; however, they neglect the correlations amid features. Conversely, wrapper approaches could not be applied due to their complexities. Moreover, in particular, existing feature selection methods could not handle such data, which is one of the major causes of instability of feature selection.

To overcome such issues, a majority voting-based hybrid feature selection method, namely feature distillation and accumulated selection (FDAS), is proposed to investigate the optimal subset of relevant features for analyzing the publisher's fraudulent conduct. FDAS works in two phases: (1) feature distillation, where significant features from standard filter and wrapper feature selection methods are obtained using majority voting; (2) accumulated selection, where we enumerated an accumulated evaluation of relevant feature subset to search for an optimal feature subset using effective machine learning (ML) models.

Empirical results prove enhanced classification performance with proposed features in average precision, recall, f1-score and AUC in publisher identification and classification.

The FDAS is evaluated on FDMA2012 user-click data and nine other benchmark datasets to gauge its generalizing characteristics, first, considering original features, second, with relevant feature subsets selected by feature selection (FS) methods, third, with optimal feature subset obtained by the proposed approach. ANOVA significance test is conducted to demonstrate significant differences between independent features.

In a multi-stakeholder recommender system (MSRS), stakeholders are the multiple entities (consumer, producer, system, etc.) benefited by the generated recommendations. Traditionally, the exclusive focus on only a single stakeholders' (for example, only consumer or end-user) preferences obscured the welfare of the others. Two major challenges are encountered while incorporating the multiple stakeholders' perspectives in MSRS: designing a dedicated utility function for each stakeholder and optimizing their utility without hurting others. This paper proposes multiple utility functions for different stakeholders and optimizes these functions for generating balanced, personalized recommendations for each stakeholder.

The proposed methodology considers four valid stakeholders user, producer, cast and recommender system from the multi-stakeholder recommender setting and builds dedicated utility functions. The utility function for users incorporates enhanced side-information-based similarity computation for utility count. Similarly, to improve the utility gain, the authors design new utility functions for producer, star-cast and system to incorporate long-tail and diverse items in the recommendation list. Next, to balance the utility gain and generate the trade-off recommendation solution, the authors perform the evolutionary optimization of the conflicting utility functions using NSGA-II. Experimental evaluation and comparison are conducted over three benchmark data sets.

The authors observed 19.70% of average enhancement in utility gain with improved mean precision, diversity and novelty. Exposure, hit, reach and target reach metrics are substantially improved.

A new approach considers four stakeholders simultaneously with their respective utility functions and establishes the trade-off recommendation solution between conflicting utilities of the stakeholders.

The Multi-Stakeholder Recommendation System learns consumer and producer preferences to make fair and balanced recommendations. Exclusive consumer-focused studies have improved the recommendation accuracy but lack in addressing producers' priorities for promoting their diverse items to target consumers, resulting in minimal utility gain for producers. These techniques also neglect latent and implicit stakeholders' preferences across item categories. Hence, this study proposes a personalized diversity-based optimized multi-stakeholder recommendation system by developing the deep learning-based diversity personalization model and establishing the trade-off relationship among stakeholders.

The proposed methodology develops the deep autoencoder-based diversity personalization model to investigate the producers' latent interest in diversity. Next, this work builds the personalized diversity-based objective function by evaluating the diversity distribution of producers' preferences in different item categories. Next, this work builds the multi-stakeholder, multi-objective evolutionary algorithm to establish the accuracy-diversity trade-off among stakeholders.

The experimental and evaluation results over the Movie Lens 100K and 1M datasets demonstrate that the proposed models achieve the minimum average improvement of 40.81 and 32.67% over producers' utility and maximum improvement of 7.74 and 9.75% over the consumers' utility and successfully deliver the trade-off recommendations.

The proposed algorithm for measuring and personalizing producers' diversity-based preferences improves producers' exposure and reach to various users. Additionally, the trade-off recommendation solution generated by the proposed model ensures a balanced enhancement in both consumer and producer utilities.