Comparative Analysis of Optimizer Methods for Machine Learning Algorithm using Search Keyword Ad Data

Authors

  • Hyunhee Park
  • Minjun Ji
  • . .

DOI:

https://doi.org/10.14419/ijet.v7i4.39.23694

Published:

2018-12-13

Keywords:

Web-search-word, Keyword advertisement, Machine learning algorithm, Optimizer mechanism, ANN, Regression

Abstract

A searching for specific keywords through an online web searching platform (e.g., Google or Naver) is one of the most popular search mechanisms. Therefore, an advertisement on the online web searching platform has become one of the representative advertising marketing mechanisms. In order to advertise a specific keyword, it is necessary to pay the web-search-word to the online web searching platform and participate in the bid system. However, since the ads of online web searching platform are operated privately, it is quite difficult to know the bidding price of specific keywords. In this paper, we compare analysis results of machine learning algorithms with various optimizers to find the targeted rank of specific keywords and the desired ranking on the online web searching platform by using the machine learning algorithms. Particularly, it is quite important to find an appropriate optimization mechanism for the machine learning algorithm because it can derive different results of the applied machine learning algorithm according to the optimization mechanism. Therefore, we propose an appropriate machine learning algorithm with various optimizers by analyzing the web-search-word advertisement data. The ANN of deep learning and regression (i.e., linear, logistic, softmax regressions) algorithms are applied for the machine learning algorithms. In addition, we applied the optimizer mechanisms of Adam, Adagrad, Gradient Descent, Momentum and RMSProp to these algorithms. Extensive simulation results show that the Adam and Adagrad optimizer mechanisms have high test accuracy rate. Specifically, it can be seen that each optimizer mechanism shows quite difference in accuracy rate according to learning rate. Finally, it is necessary to analyze the machine learning algorithm applying various optimizer mechanisms to present the bidding price prediction model of the web-search-word advertisement. In this paper, it makes it possible to predict the optimal bidding price for the web-search-word advertisement.

 

 

References

[1] Hou, L. (2015). A Hierarchical Bayesian Network-Based Approach to Keyword Auction, IEEE Trans. on Engineering Management, 62(2), 217–225.

[2] Lauritzen, S. (1995) The EM algorithm for graphical association models with missing data, Comput. Statistics Data Anal., 19(2), 191–201.

[3] Jansen, B.J., Zhang, M., & Schultz, C. D. (2009). Brand and its effect on user perception of search engine performance, Journal of the association for information science and technology, 60(8), 1572–1595.

[4] Auerbach, J., Galenson, J., & Sundararajan, M. (2008). An empirical analysis of return on investment maximization in sponsored search auctions, Las Vegas, NV, USA: International Workshop Data Mining Audience Intell. Ad.

[5] Jerath, K., Ma, L., Park, Y., & Srinivasan, K. (2011). A Position Paradox in sponsored search auctions, Marketing Sci., 30(4), 612–627.

[6] Kingma,D.&Ba,J. (2008). ADAM: A Method for Stochastic Optimization,San Diego, CA, USA: International Conference on Learning Representations.

[7] Duchi, J., Hazan, E., & Singer Y. (2011). Adaptive subgradient methods for online learning and stochastic optimization, The Journal of Machine Learning Research, 12(2011), 2121–2159.

View Full Article:

How to Cite

Park, H., Ji, M., & ., . (2018). Comparative Analysis of Optimizer Methods for Machine Learning Algorithm using Search Keyword Ad Data. International Journal of Engineering & Technology, 7(4.39), 30–36. https://doi.org/10.14419/ijet.v7i4.39.23694

License

Authors who publish with this journal agree to the following terms:

    1. Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution Licensethat allows others to share the work with an acknowledgement of the work''s authorship and initial publication in this journal.

    1. Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal''s published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.

  1. Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
Received 2018-12-11
Accepted 2018-12-11
Published 2018-12-13