| Summary: | 碩士 === 國立清華大學 === 資訊工程學系 === 101 === Consuming activities is one of the most interesting behaviors for researchers, which is
major profitable source to online web service. As smartphone have become prevalent and
have the ability to sense user context, mobile offers a great opportunity to understand
user consuming intention. In this thesis, we present a next consuming behavior prediction
framework for smartphone users. We predict user next possible purchase item by collecting
useful context from smartphone and extract semantic context for further study. As one of
the key enabling techniques, a probabilistic prediction model has proposed to better describe
user consuming behavior. To demonstrate the feasibility of proposed framework, we evaluate
the overall framework by constructing a context collection daemon in Android and asking
14 participants conduct a 3 weeks experiment by using Easycard in every transaction during
daily life. The result indicates that timing and location are the most important context
for next consuming activity prediction and our framework reach 76% accuracy in overall
evaluation.
1
|
|---|
