Event-Integrated Forecasting Model for Call Volume Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Historical data-based forecasting in call centers and similar entities often fails to account for current events and intelligence from speech and text analytics, leading to inaccurate demand predictions.
Innovation Solution
A system that records event data alongside call volume information, trains models to predict demand using historical and event data, and identifies external events affecting demand, allowing for more accurate scheduling of agents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If historical data alone is used for forecasting, then the forecasting system is simple and easy to operate, but the forecast accuracy deteriorates because current events and external factors are not accounted for
Solution Approach 1:
The patent combines historical demand data with event data (weather, sports, TV shows, political events) into a unified forecasting model. The system merges multiple data sources and multiple forecasting models (baseline model using historical data only, and enhanced models incorporating event data) to achieve more accurate predictions while maintaining manageable complexity through structured integration.
Solution Approach 2:
The forecasting system is segmented into separate components: a baseline forecasting model that uses only historical data, and enhanced forecasting models that incorporate event data. This segmentation allows the system to handle complexity by processing different types of data through separate models and then combining their insights, making the overall system more manageable while improving accuracy.
2Measurement precision
If event data is incorporated into forecasting models, then forecast accuracy improves, but the difficulty of detecting and measuring relevant events increases
Solution Approach 1:
The patent introduces an intermediary layer of event data collection and processing that bridges the gap between raw event information and the forecasting models. This intermediary system collects, cleans, and formats event data from multiple sources, making it ready for model consumption and reducing the direct complexity of detecting and measuring relevant events.
Solution Approach 2:
The system employs a universal event data collection framework that can handle multiple types of events (weather, sports, TV shows, political events) through a consistent process. This multi-functional approach simplifies event detection by using the same mechanisms for different event types, reducing the overall complexity despite the diversity of data sources.
3Adaptability or versatility
If multiple data sources are integrated, then the adaptability to changing conditions improves, but the device complexity increases
Solution Approach 1:
The forecasting system is designed to be dynamic, allowing it to adapt to changing conditions by incorporating real-time event data alongside historical patterns. The system can adjust its behavior based on the type of event occurring (weather changes, sports events, TV shows, political events) and automatically modify forecasts accordingly, providing high adaptability while managing complexity through automated adjustment mechanisms.
Data Source
AI summary
In an entity such as a call center, back office, or retail operation, external event data is recorded along with call volume information for a plurality of time intervals. Based on the recorded event data and call volume for the plurality of intervals, a model is trained to predict call (or other communication) volume for a specified time interval using the external event data. The external event data may include data about one or more events that may affect the demand received by the entity. When the predicted call volume is significantly above or below what would be predicted for the entity using historical data alone, an indicator may be displayed to a user or administrator that identifies the external event that is responsible for the lower or higher prediction. The call volume prediction may be used to schedule one or more agents (or other employees) to work during the specified time interval.