HYBRID MODEL BASED APPROACH FOR ACCURACY IN ADVERTISEMENT VIEWABILITY PREDICTION BASED ON USER AND PAGE FEATURES

Author (s) : M.Ushanandhini and Dr.C.Viji

---

Abstract :
Advertisement is a vital source for products and its branding. Nowadays online advertisements became an emerging field for producers and also online publishers. Prediction of user’s interest and choices are more important to place what kind of advertisements placed in the web page to get more number of views and clicks.This work proposed a hybrid model using probabilistic latent class model and logistic regression classifier. The process performed in two stages. Probabilistic Latent Class(PLC) model used to predict the advertisement viewability prediction in the first stage. In the existing works, web page based features especially scroll depth and browser behaviour are used in the prediction process. Halt time and its level, scroll depth level, gender, type of advertisement, viewed status features are additionally used in the proposed work to predict the viewabilty in using PLC model.In the second stage, Logistic Regression classifier used to classify the collected advertisement view data set and find the accuracy in viewabilty prediction. Predicted data set is collected from the database and perform the classification process using Logistic regression.The stage 1 is implemented using .NET Technology with MS-SQL Server Databaseand stage 2 is implemented in Weka tool. This hybrid approach improves the accuracy inprediction of advertisement viewability.

Keywords : Adview Prediction, Viewers Opinion, PLC, Logistic Regression

Conference Date : 30/08/2019

Pages : 11

Paper ID : 19011