PSAC: Context-Based Purchase Prediction Framework via User's Sequential Actions
Along with the daily operation of e-commerce web services, a sig-nificant quantity of data has been recorded. The research of user’sbehaviors based on the collected data has generated intense atten-tion for accurately offering services that can match the customer’sneeds and predict the purchase actions. Traditionally, most of theresearches utilize only the behavioral instances between users andproducts, i.e., browse or click history, and session status. However,these features provide only a fundamental knowledge of the givenuser rather than the rationale behind their actions. We find thatquery should play an important role as well as it is the main entrypoint for users when arriving e-commerce website. Since usersutilize queries to decide the direction of succeeding event, the se-mantic meanings of these queries demonstrate a particular linkwith the action.In this paper, we propose thePrediction framework that ana-lyzes User’sSequentialActions viaContext (PSAC) to exploit theconnection between the user’s searching keywords and behaviorsto investigate their ultimate intention on an e-commerce websiteand improve the purchase prediction accuracy. We utilize the e-commerce dataset provided by Yahoo Taiwan, one of the largest webservices provider in Taiwan. According to our preliminary analysis,we design a session-based structure to deal with the environment-shifting (influenced by coexisting fashion), and experience-shifting(changed through user’s actions) issues which we observed in thedataset. In the simulation section, we apply two deep learningframeworks to perform the prediction task. Experimental resultsconfirm that queries serve as a critical matter in perceiving a user’spurchasing intention. Moreover, the proposed framework could sig-nificantly improve the prediction accuracy compared with baselinemethods.
