Predictive Behavior Model Using Unsupervised Feature Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for predicting user behavior require human intervention and are susceptible to noise and randomness in large data sets, failing to accurately account for temporal user behavior, leading to inaccurate predictions.
Innovation Solution
A method using unsupervised feature learning and a deep recurrent neural network, specifically connecting deep restricted Boltzmann machines to encode user behavior data into feature representations, which are then used to generate a predictive model that predicts user behavior without human intervention and is less susceptible to noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human intervention is used to determine features representing user behavior, then the system can interpret behavior patterns, but the system becomes susceptible to noise and randomness in large data sets and produces inaccurate predictions
Solution Approach 1:
The patent replaces manual feature determination (mechanical/human process) with an automated machine learning system that uses unsupervised feature learning and deep recurrent neural networks to extract behavior features from raw data, eliminating human intervention while improving prediction accuracy through automated pattern recognition
Solution Approach 2:
The system performs self-service by automatically learning features from user behavior data without human intervention. The deep recurrent neural network autonomously identifies patterns and generates predictions, making the system self-sufficient in feature determination while handling large data sets effectively
2Reliability
If existing machine-learning systems analyze large data sets to predict user behavior, then comprehensive behavior patterns can be captured, but the systems fail to account for temporal user behavior and produce inaccurate predictions
Solution Approach 1:
The patent employs a dynamic recurrent neural network architecture that adapts to temporal variations in user behavior. The network dynamically processes sequences of user actions over time, adjusting its internal state to capture evolving behavior patterns, thereby achieving both reliability and adaptability for temporal data
3Measurement precision
If deep recurrent neural networks with unsupervised feature learning are used, then accurate prediction of user behavior is achieved, but the system complexity increases
Solution Approach 1:
The patent segments the complex prediction task into distinct components: an unsupervised feature learning module that extracts behavior features, a deep recurrent neural network module that processes temporal sequences, and a prediction module that generates outcomes. This segmentation manages complexity by organizing functions into modular, independently trainable components
Data Source
AI summary
Certain embodiments involve a model for predicting user behavior. For example, a system accesses user behavior data indicating various users' behaviors during intervals over various periods of time and target behavior data indicating a particular user behavior. The system associates each user with a label that indicates whether a user performed a particular action during or after a time period based on the target behavior data. The system uses the user behavior data to train various deep Restricted Boltzmann Machines (“RBM”) to generate representations of each user over each period of time that indicate the user behavior over the time period. The system generates a predictive model by connecting the RBMs into a deep recurrent neural network and uses the target behavior data associated with each user, along with the representations of each user, as input data to train the deep recurrent neural network to predict user behavior.