Study On Mining Methods And Implementations Of Smartphone APP Logs For Understanding User Behaviors

As the main intelligent mobile terminal device,the smartphone has become an indispensable part of people's daily life.With its powerful functions,smartphones have occupied the vast majority of the mobile phone market share.Differ from traditional mobile phones,smart phones are equipped with smart operating systems,embedded with numbers of sensor with different functions.In particular,smartphones also support the manual selfinstallation and uninstallation of third-party applications(APPs)themselves.These featurerich APPs of smartphones with good user interfaces become one of the important factors to attract users.Therefore,APP developers always pay attention to users' needs for different APPs,and at the same time,they also capture the changes trends in these needs through different channels in order to provide convenient services that best meet the user needs.Therefore,the number and types of APPs becomes more and more abundant.The use of APP has been integrated into the daily life of users,from online learning,online shopping,roll-in check-in,office interaction,news browsing,leisure and entertainment,etc.These behaviors of users will be recorded in the form of APP logs.Since APP log is a true record of users behavior,mining and analysis of an individual's historical APP log can extract the personal sacrifice information of behavior patterns and personality preferences hidden behind the data.With the goal of mining users behavior patterns and extracting users preferences,this dissertation studies the mining and analysis of APP logs data,as follows:1)Collection of smart phone APP logsThis dissertation is based on the Android platform.First,an app state data collection tool was developed.Used to collect the user's start time,duration and APP package name.The research in this area is based on the capture and analysis of user behaviors based on the smartphone usage.The usage of mobile APPs reflects the user's habits and behaviors to a certain extent.The accumulated huge amount of historical APP log data is a true record of user usage and is an user.An important source of information for behavior pattern mining and extraction.To this end,this dissertation studies the continuous collection of APP usage data.2)APP log data preprocessing and formalizationPre-processing and formalize of the obtained APP usage data.The APP log is a record of the APP running status and recording the user's usage of the app.When mining user usage patterns based on APP logs,it is necessary to filter out APP short running records and desktop running records that do not reflect user usage in the original APP log,meanwhile filling in the lack of APP running due to device or program fall recording.Formally,the pre-processed APP usage data represents a triple with time labels to form an APP transaction set.3)APP transaction association rule extraction and discoveryThe classic data mining algorithm Apriori is used to mine the association rules of the collected APP usage data to obtain associations between frequently used APPs and analyze the users behavior habits,hobbies,and personality characteristics in using smartphones.The association between APP transactions reflects their concurrency.The association rules of APP transactions are a reflection of users behavior patterns,and provide a technical implementation basis for the implementation of services such as reminders and recommendations.4)Design and implementation of APP log mining systemBased on the aforementioned algorithms and technologies,as well as using advanced software engineering ideas,theories,and methods,a lightweight APP log mining system that extracts users behavior patterns is designed and implemented.At the same time,the system is tested in real application scenarios.By integrateing all the historical data used by personal APPs,and applying reasonable data mining techniques to discover and identify user behaviors,interests,hobbies,learning conditions,occupation types as well as consumption levels hidden in APP data to provide personalization Suggestions and reminders,while improving user experience and reducing or eliminating the negative impact brought by the use of mobile phones,have a research significance and application value.