How Daynamica works
Daynamica™ combines GPS sensing with advanced statistical and machine learning techniques to automatically detect, identify, and summarize attributes of daily activity and travel episodes. It then allows users to view, correct, and provide additional information on the activity and travel episodes at their convenience— every day, at home or on the go.
Data from the two sources—the sensors and the users—interact to inform, calibrate, and augment each other:
- Sensor data are analyzed and processed locally on the smartphone in real time to extract meaningful activity and travel information. The extracted information serves as a basis to prompt the user for more detailed and more accurate information.
- The user-entered data in turn optimize how the sensor data are analyzed and processed. This increases the accuracy of the information extracted from sensor data over time—the algorithm “learns” about a user’s routine travel locations and behaviors—so that the user needs to make fewer updates or corrections.
Performance Test Results
A series of laboratory tests and two rounds of seven-day field tests found that the app has:
- Reasonable battery consumption rate. With Daynamica running continuously, 74% of the phones had a battery life longer than six hours, and about half the phones had a battery life longer than eight hours.
- Moderate data storage and transmission requirements. Daynamica produces 50 megabytes (uncompressed) of data per day; the associated weekly data transfer needs are roughly 150 megabytes after data compression.
- Good accuracy. Daynamica had high accuracy in identifying activity versus trip episodes (90%) and in classifying the travel modes of each trip episode (86%), and medium-high accuracy in classifying the types of activity episodes (user-specified activity type is among the top two most probable predicted activity types 70-80% of the time).