With people using smart phones more and more, the number of mobile apps is increasing exponentially. Users may install dozens of apps on their phones, and it is becoming a challenge for them to easily find the app that they want to use. In our research, we studied how to accurately predict the next app that the user will use, depending on the context. Our approach/methodology consisted of three phases. Firstly, the data was preprocessed; noisy data was removed (e.g. system apps, and apps that aren’t found in Play Store) and contextual data (e.g. movement, location, and time) was generated from raw data (e.g. longitude, latitude, and timestamp). Secondly, the data was analyzed in terms of app usage and distribution of app usage in time based on different contexts. As last step, we created a hybrid algorithm (mainly involved with Most Frequently Used – MFU – algorithm, Naïve Bayesian) to perform better than the state-of-art techniques. The algorithms run under the Python Spark framework. The data represents a week’s events from Yahoo Aviate Launcher.
La presenza dello smartphone, oggigiorno, si sta rendendo sempre più nota, e in conseguenza il numero delle applicazioni aumenta esponenzialmente. Gli utenti possono installare dozzine di applicazioni, e, quindi, trovare facilmente quella desiderata diventa difficile. In questo articolo, abbiamo studiato un metodo per predire accuratamente la prossima applicazione che l’utente userà, in base al contesto. La nostra ricerca si è composta di tre fasi. Nella prima parte, i dati sono stati pre-processati: i dati irrelevanti sono stati esclusi (app del sistema e app che non sono presenti nel Play Store) e sono stati generati dati contestuali (es. movimento, posizione, e orario) a partire dai dati grezzi (es. longitudine, latitudine, e marcatura temporale). In secondo luogo, i dati sono stati analizzati, in termini di uso delle app e della distribuzione nel tempo di questo uso, in base a diversi contesti. Infine, un algoritmo ibrido è stato creato (usando l’algoritmo Most Frequently Used – MFU – e l’algoritmo Naïve Bayesian) per ottenere migliori risultati delle tecniche attuali. Gli algoritmi operano sotto il framework Python Spark. I dati usati sono gli eventi di una settimana, presi dal Yahoo Aviate Launcher.
Mobile apps recommendation
ERGUN, MERT;CHEN, YUXING
2015/2016
Abstract
With people using smart phones more and more, the number of mobile apps is increasing exponentially. Users may install dozens of apps on their phones, and it is becoming a challenge for them to easily find the app that they want to use. In our research, we studied how to accurately predict the next app that the user will use, depending on the context. Our approach/methodology consisted of three phases. Firstly, the data was preprocessed; noisy data was removed (e.g. system apps, and apps that aren’t found in Play Store) and contextual data (e.g. movement, location, and time) was generated from raw data (e.g. longitude, latitude, and timestamp). Secondly, the data was analyzed in terms of app usage and distribution of app usage in time based on different contexts. As last step, we created a hybrid algorithm (mainly involved with Most Frequently Used – MFU – algorithm, Naïve Bayesian) to perform better than the state-of-art techniques. The algorithms run under the Python Spark framework. The data represents a week’s events from Yahoo Aviate Launcher.File | Dimensione | Formato | |
---|---|---|---|
thesis.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Thesis text
Dimensione
1.13 MB
Formato
Adobe PDF
|
1.13 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/131901