APP CLASSIFIER

Idea behind project is to solve problem of people who are downloading unnecessary apps from play store without having any prior information. The system is for user to get a detailed classification of apps relevant to his/her category. 

Need of Classification of Applications 

Presently, in Google app-store classification of apps are done is in single level hierarchy. So user can’t appropriately know about the classification of apps. So we have developed project which classifies apps in multilevel hierarchy. Due to such classification it becomes easy to user to find apps as per the need of user. And classification also helps to recommend apps to user as per user’s occupation, age, or gender.

For the App usage analysis the major step is to classify Apps into some categories which are predefined. However, it is difficult task to clearly classify mobile Apps due to the limited contextual information available for the analysis. For example, a limited contextual information about mobile Apps is present in their names. However, this contextual information is usually incomplete and ambiguous. We propose an approach for first extracting the contextual information of mobile Apps by exploiting the additional Web knowledge from the Web search engine. Then, we also extract some contextual features for mobile Apps from the context-rich device logs of mobile users. Finally, combining all the enriched contextual information into the Maximum Entropy model for training a mobile App classifier. Finally, we combine all the enriched contextual information into the Maximum Entropy model for training a mobile App classifier. To validate the proposed method, we conduct extensive experiments on mobile user’s device logs to show both the effectiveness and efficiency of the proposed approach.

The proposed approach for mobile App classification consists of two main stages. First, we collect many context logs from mobile users, and extract both Web knowledge based features and real-world contextual features for the Apps appearing in these logs. Second, we take advantage of the machine learning model for training an App classifier. To be specific, given a target taxonomy, an App A and a system-specified parameter K, our approach incorporates the features extracted from both the relevant Web knowledge and contextual information of A to classify A into a ranked list of K categories. 

A critical problem of this type of features is that the lengths of App names are usually too short and the contained words are very sparse. As a result, it is difficult to train an effective classifier by only taking advantage of the words in App names. Moreover, the available training data are usually with limited size and may not cover a sufficiently large set of words for reflecting the relevance between Apps and category labels. Therefore, a new App who’s partial, or all words in name do not appear in the training data will not obtain accurate classification results if the classifier is only based on the words in App names. To this end, we should also take consideration of other effective features which can capture the relevance between Apps and category labels. In the following section, we introduce the features extracted from both the relevant Web knowledge and real-world contextual information for training an App classifier.

Hardware Requirements

  • Processor
  • Min core- i3
  • RAM Minimum 2GB
  • Hard Disk 40GB
  • Android Emulator for testing 

Software Requirements 

OS  Android (Ice cream Sandwich 4.0) and above versions, Windows, Linux – WAMP / XAMPP For MySql  Apache Tomcat 7.0.56  For JSP / JAVA servlets