More industries are introducing core technologies and services that can analyze user and computer interactions to improve personal work performance and productivity, as well as monitoring of employees who work from home under COVID-19 pandemic era.
Nexus, a law firm requested the development of AI-based user interaction analysis solution to replace existing manual workflow management system and utilize AI technology to recommend specific work steps based on analyzation of work patterns of individuals to optimize workflow and efficiency.
User interaction data collection
Interaction data
collection agent
Time-series data processing
Data Culstering
based on tasks
Time report generation
Technological Challenges
Automation of workflow by automatically accumulate and analyze user’s PC data without user’s command, provides real-time analysis and prediction of work patterns and visualize user’s hourly processes based on interaction data.
User interaction data collecting & processing platform
To increase the utilization of the data collected by the existing Window’s interaction data collection agents, data format integration and format standardization process are required. It is required to develop data collection agents comprise of system information, process information, mouse event information, keystroke information and screen capture data.
1
Time-series data clustering / workflow analysis-based process recommendation
Extended application of time series clustering algorithm on PC interaction data through wide tests and application of traditional clustering algorithm expansion-based technology such as IDEC(Improved Deep Embedded Clustering), DTC (Deep Temporal Clustering), SOM-VAE, N2D, TSC-CNN and DBSCAN, k-shape, USSL analysis to input error correction, identify possible automatable points
2
Dashboard & Visualization
Visualization of user productivity + work process recommendation and correction notification
3
Road Map
Development of interaction collection model
User/task-based data clustering
User/manager dashboard
Time report generation
Key Features
User interaction data collection: Keyboard stroke, mouse coordinates, screen capture
1
Input error correction recommendation
2
Workflow based possible tasks recommendation
3
User identification based on interaction data
4
User time report generation
5
The Result
95%
input error collection & recommendation accuracy
95%
User identification accuracy
2%
less than
PC performance degradation due to
user data collection & analysis process
90%
above
User identification accuracy