Can An App's Characteristics Predict Its Success?

Abstract

On the Google Play Store, applications have a variety of characteristics, from category of content and cost to version and rating. In this paper, we analyze the relationship between the success of a Google Play Store app (as determined by an app’s rating out of five and its review count) and its characteristics: content category, price, and amount of installs. By using linear regression, random forest, and multilayer perceptron models, we found that an app’s install count, whether or not it cost money, and the type of content all have a significant effect on its success. During modeling, we found that the random forest model was the most successful, with a training and testing RMSE of 0.079 and 0.081, respectively, and a training and testing R2 of 0.185 and 0.180, respectively. Based off these results, we have confirmed that there is a correlation between success and category of content, install count, and price. The results from this paper can inform app developers and investors about optimal statistics for the most successful applications.