Abstract

This project is to build a model to predict whether a user will download an app after clicking a mobile app advertisement, based on a vast collection of users’ clicking record on app advertisement as our original training data including attributes such as ip, os, device, channel etc. Recent years have witnessed an inevitable phenomenon where more and more users simply click on the ad link of apps without downloading them, which results in considerable wasted ad investment from app service providers. Therefore, it is worthwhile for us to build a model so that such problem could be solved or its latency could be minimized to utmost extent.

Among the given attributes, ‘is_attributed’ and ‘attributed_time’ are chosen as output labels in training process and as predicted result in testing process. Some approaches investigated in this project include Decision Tree (DT), Random Forest (RF), Ada Boosting (AB), K-Nearest Neighbor (KNN), Logit Boosting (LB), and Multilayer Perceptron (MLP). We initially tried and tested some of these algorithms on Weka, where we observed the results of DT, RF, DT, and then we implemented MLP regressor & classifier using sklearn packet in Python. The best prediction accuracy is approximately 95% when training on MLP regressor & classifier.