Predicting click-through rates using data mining technique on digital advertisements

Abstract

In times of increasingly busy use of social media, placing advertisements on social media such as Facebook is an attractive alternative. With the various advantages of advertising on social media, making it very suitable for MSMEs. The use of video as a format for delivering advertising messages is also rife because of the faster internet speed. However, the cost of making ads in the form of videos is relatively more expensive so it needs a lot of consideration when making it be efficient. One thing that advertisers often pay attention to on social media is the click-through rate. This variable becomes one of the measures of the effectiveness of an advertisement. Hence predicting click-through rate is become important nowadays for advertisers, especially to those who have budget constraints. This research tries to predict the click-through rate using data mining techniques. This paper use CRISP method. The dataset was taken from a Facebook advertisement from a small-medium enterprise in Indonesia. Video watches at 25%, 50%, 75%,95% and 100% is use as predictors. The results show that data mining can be used to predict the click-through rate using video watches percentage. Deep learning is the most suitable model for this prediction. The interpretation of the results from data mining is done and managed to find the variables that support the predictions and contradict the predictions.