Publications
Mining App Stores: Extracting Technical, Business and Customer Rating Information for Analysis and Prediction (Under Review)
*** This recent work is currently under review. In its present form it provides only figures from our analysis. However, if the paper is accepted then this section of the website will extended. It will be populated with data from the paper to support replication. More details can be found in our technical report.
App development is an increasingly innovative and lucrative software industry. However, determining a suitable market price of an App is both demanding and critical; the comparatively low unit price, but considerable volume of sales dramatically increases the impact of miss-pricing. In this paper we leverage app store repository mining and machine learning, to automatically construct predictive models for this prediction problem. We implement and evaluate our approach on 9,588 non-free Apps from the Blackberry App Store, demonstrating that our approach statistically significantly outperforms existing approaches with at least medium effect size in 15 out of 17 (88%) of Blackberry App Store categories.
Results
Comparing the MAR values provided by CBR (best, worst and mean results) and Random Guessing
Comparing the MAR values provided by CBR (best, worst and mean results) and current marketing price strategies
