Operation Log Classification Using Similarity-Based Work Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods struggle to easily and accurately classify a series of operations from start to end, making it difficult to grasp the type and amount of work performed by individuals, which hinders effective business improvement measures such as robotic process automation (RPA).
Innovation Solution
A classification device and method that utilize an identification unit to acquire operation logs, create vectors based on co-occurrence relationships, calculate similarities between adjacent operations, determine division points, and classify operation sets into classes based on common types of operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If operation logs are analyzed in fine granularity units of individual operations, then detailed operation information is obtained, but it takes time and effort to grasp actual operations in units of work including series of operations
Solution Approach 1:
The patent segments the continuous operation log into discrete operation units by detecting division points where operation types change. Each operation unit contains a series of operations between two division points, representing a complete work task. This segmentation enables efficient classification while preserving detailed operation information.
Solution Approach 2:
The patent transforms the operation classification problem from analyzing individual operations to analyzing operation sequences in a temporal dimension. By creating operation units that span multiple operations and classifying them as whole entities, the system efficiently grasps work-level information without losing operational details.
2Productivity
If operation logs are analyzed in units of work including series of operations, then actual work information is grasped efficiently, but it becomes difficult to accurately classify groups of works from start to end
Solution Approach 1:
The patent uses the operation type at each division point as feedback to determine work boundaries. The detection of operation type changes provides feedback signals that automatically identify where one work unit ends and another begins, enabling accurate classification without manual intervention.
Solution Approach 2:
The patent performs preliminary classification by detecting operation types and identifying division points before final work classification. This preliminary action of segmenting the operation log into operation units with clear boundaries facilitates subsequent accurate classification of complete work groups.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A classification device (10) acquires an operation log related to operation information, identifies each operation performed by a user using the operation log, and creates a vector of each operation on the basis of a co-occurrence relationship between the identified operations. Then, the classification device (10) calculates a similarity between a predetermined number of operations adjacent to each other in chronological order using the created vector of each operation, determines a division point of the operations using the calculated similarity, and divides a time-series operation into operation sets on the basis of the division point. Then, the classification device (10) classifies the operation sets into classes on the basis of the number of types of operations common to the operation sets.