Time Series Segmentation for Cost Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Forecasting long-term costs for service delivery projects is complex due to the involvement of multiple variables and requires sizable amounts of input data, making existing methods inefficient for accurate cost estimation.
Innovation Solution
A system and method that generate a succinct approximate representation of a time series by partitioning it into segments based on peak and trough data values, creating a sequence of segments that rise and fall alternately, and calculating totals for each segment to represent the input series, allowing for better cost forecasting and project similarity analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional cost forecasting methods are used for service delivery projects, then comprehensive cost estimation can be achieved by considering multiple variables, but the process becomes complex and requires sizable amounts of input data
Solution Approach 1:
The patent segments the time series data into multiple segments based on peak and trough values, transforming a complex continuous forecasting problem into discrete segment-based analysis. Each segment represents a distinct phase (rising or falling) that can be analyzed independently, reducing overall system complexity while maintaining forecasting accuracy
Solution Approach 2:
The patent transforms the original time series data into a new representation by identifying peak and trough values and creating segments based on these parameter changes. This parameter transformation simplifies the data structure from continuous values to discrete segments with specific characteristics (rising/falling), making the forecasting process more manageable
2Measurement precision
If detailed time series data is analyzed for cost forecasting, then accurate cost estimation is achieved, but the processing time and computational resources increase
Solution Approach 1:
The patent extracts only the essential features from the detailed time series data - specifically the peak and trough values - to create a simplified representation. By taking out only these critical data points and forming segments around them, the method maintains forecasting accuracy while dramatically reducing the amount of data that needs to be processed
Solution Approach 2:
The time series is divided into segments that group multiple data points between peaks and troughs. This segmentation allows the system to process one segment at a time rather than analyzing every individual data point, improving processing efficiency while preserving the essential cost patterns
Data Source
AI summary
Embodiments of the present invention provide a system, method and computer program product for generating a succinct approximate representation of a time series. A method comprises determining at least one peak data value and at least one trough data value of an input series comprising a sequence of data values over time. The input series is partitioned into multiple segments comprising at least one rising segment that rises to a peak data value and at least one falling segment that falls to a trough data value. A sequence of segments that rise and fall alternately is generated based on the segments. A sequence of totals representing a succinct approximate representation of the input series is generated. Each total comprises a sum of data values for a corresponding segment of the sequence of segments.