Unmet Need: Simple and reliable approach to activity prediction
An individual’s activities affect that individual, society, and the environment. Over the past decade, maturing of data mining and pervasive computing technologies have made it possible to automate activity learning from sensor data. To observe and understand individuals' activities and their effects, and predict activities that will occur in future, baseline methods exist, but are not reliable and accurate.
The Technology: Activity predictor in a prompting application
This invention describes an Activity Predictor, the goal of which is to predict future activity occurrence times from sensor data, and introduce a data-driven method for performing activity prediction. The activity prediction problem is formulated as imitation learning framework and provided as simple regression learning. Performance metrics desirable for the algorithm’s performance are selected in context of real-world applications. The learned predictor is embedded into a mobile device based activity prompter application and evaluated on multiple participants living in smart homes.
• Activity-aware services such as energy-efficient home automation, prompting-based interventions, and anomaly detection.
• Security Systems and computer games.
• Simple and reliable approach to prediction of activities from sensor data.
• Re-usability and re-adaption of application based on new user data.
• Prediction of future activities in minutes of actual occurrence.
Patent Pending, Application Published US 2017/0109656