Call Center Agent Anomaly Detection via Functional Data Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing unusual patterns in call center data, such as high call rate bursts, fail to account for inherent relationships within time series data, limiting their ability to identify anomalies effectively.
Innovation Solution
A method involving normalization of call center agent data curves, determination of robust standard deviations, and identification of outliers using threshold values to flag unusual agent behavior, which involves receiving and processing functional data to create normalized points, determining variations, and identifying points exceeding threshold values as outliers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional outlier detection methods are used to identify unusual behavior in call center data, then the process is simple and quick, but the ability to address inherent relationships within time series data is lost
Solution Approach 1:
The patent transforms time series data into functional data by changing the parameter representation from discrete time points to continuous functions. This involves representing call center metrics (call rates, handle times, abandonment rates) as continuous curves over time, allowing for more sophisticated anomaly detection that captures temporal relationships and patterns that traditional discrete methods miss.
Solution Approach 2:
The patent adds a functional dimension to the analysis by treating time series data as functions rather than sequences of points. This dimensional transformation enables the use of functional data analysis techniques that can detect anomalies based on the shape, slope, and other characteristics of the functional representations, providing deeper insights into unusual patterns.
2Measurement precision
If functional data analysis with normalization and thresholding is applied to identify unusual patterns, then anomaly detection precision improves, but computational complexity and processing time increase
Solution Approach 1:
The patent applies normalization and baseline establishment as preliminary actions before performing anomaly detection. By pre-processing the functional data to establish expected patterns and thresholds in advance, the system reduces the computational burden during real-time analysis, allowing for faster detection without sacrificing accuracy.
Solution Approach 2:
The patent replaces complex mechanical computation with statistical and functional analysis methods. Instead of using resource-intensive machine learning models, the system employs functional data analysis with normalization, variation calculation, and threshold comparison, which are computationally more efficient while maintaining high detection accuracy.
Data Source
AI summary
A method of managing a call center that includes a plurality of agents may include receiving functional data associated with the plurality of call center agents over a time period, for each data curve, normalizing the data points to create a plurality of normalized functional data points, determining a variation associated with each normalized functional data point, determining one or more threshold values associated with each normalized functional data point, determining whether one or more of the normalized functional data points have a value that exceeds the associated threshold value, and identifying each of the normalized functional data points having a value that exceeds the associated threshold value as an outlier. The method may include identifying the call center agent associated with each curve that has an outlier and presenting information pertaining to one or more identified call center agents to a user.