Summarized Behavioral Prediction

Shih-Chieh Su

Abstract: In this work, we study the topical behavior in a large scale. Both the temporal and the spatial relationships of the behavior are explored with the deep learning architectures combing the recurrent neural network (RNN) and the convolutional neural network (CNN). To make the behavioral data appropriate for the spatial learning in the CNN, several reduction steps are taken in forming the topical metrics and placing them homogeneously like pixels in the images. The experimental result shows both temporal and spatial gains when compared against a multilayer perceptron (MLP) network. A new learning framework called the spatially connected convolutional networks (SCCN) is introduced to better predict the behavior.

TL;DR: This work utilizes the CNNs and the RNNs to explore the spatial-temporal relationship among the topics of the massive log data.

Keywords: Deep learning, Supervised Learning, Applications, Structured prediction

Conflicts: qualcomm.com, fico.com