Latent Dirichlet Allocation (LDA)-guided Deep Neural Network for Pattern Discovery in Clickstream Data

10/04/2017
1:30 pm - 4:00 pm
Expo Hall/Career Fair

Objective: Inspiration
Audience Level: All
Session Type: Poster

We propose an unsupervised framework for clickstream pattern discovery using Latent Dirichlet Allocation and Deep Neural Network (LDA/DNN). LDA/DNN-detected clickstream topics map meaningfully to different product features. Coupling with A/B testing, our framework can quantify the causal impact of individual A/B testing in term of changes in usage of LDA/DNN detected clickstream topics.

Speaker(s)

Tharathorn Rimchala, Data Scientist, Intuit Inc