Keyword Search Volume Forecasting via Seasonal Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for forecasting keyword search volume are time-consuming and impractical for general use, especially for search terms without meaningful usage data over multiple years, as they require extensive historical data and are computationally intensive.
Innovation Solution
A system and method that categorize keywords into seasonal and non-seasonal categories, using a seasonal correlation value to determine the type of forecast, with ARIMA forecasts for keywords with high seasonal correlation and simplified calculations for those with low correlation, incorporating category-level seasonal variation patterns to generate search volume forecasts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional forecasting methods using accumulated multi-year data are used, then forecast accuracy is improved, but time consumption and computational intensity increase significantly
Solution Approach 1:
The patent segments keywords into different categories (e.g., seasonal vs. non-seasonal, high-volume vs. low-volume) and applies different forecasting methods to each segment. This allows computationally intensive multi-year accumulation methods to be applied only where necessary, while simpler methods handle other cases, thus reducing overall time consumption while maintaining accuracy where it matters most.
Solution Approach 2:
The patent changes the parameter of historical data duration from fixed multi-year to variable based on keyword characteristics. For some keywords, it uses multi-year data; for others, it uses shorter periods or alternative data sources. This flexible parameter adjustment reduces unnecessary computational burden while maintaining forecast accuracy for each specific case.
2Reliability
If conventional forecasting methods requiring multiple years of historical data are used, then forecast reliability is improved, but applicability to new search terms deteriorates
Solution Approach 1:
The patent performs preliminary classification of keywords to identify which ones have sufficient historical data and which are new or emerging. This preliminary action enables the system to apply appropriate forecasting methods from the start, using alternative approaches for new terms that don't require multi-year data, thus maintaining both reliability for established terms and adaptability for new ones.
Solution Approach 2:
The patent introduces intermediary methods such as category-based forecasting, where new search terms are grouped into broader categories with established patterns. The category-level data serves as an intermediary that provides reliable forecasts for new terms by leveraging data from related, more established terms, thus bridging the gap between reliability and adaptability.
3Measurement precision
If advanced forecasting methods like ARIMA are applied to all keywords, then forecast accuracy for seasonal terms is improved, but computational complexity increases
Solution Approach 1:
The patent segments keywords based on their seasonal characteristics and applies ARIMA or similar advanced methods only to the seasonal segment, while using simpler methods for non-seasonal keywords. This segmentation maintains high forecast accuracy for seasonal terms while avoiding unnecessary computational complexity for terms where advanced methods provide no benefit.
Solution Approach 2:
The patent applies different levels of forecasting complexity to different keywords based on their specific characteristics. Advanced methods like ARIMA are applied locally only where needed (seasonal, high-volume keywords), while simpler methods are used elsewhere. This local quality approach optimizes the balance between accuracy and computational complexity for each specific case.
Data Source
AI summary
A method and system are provided for forecasting keyword search volume. Keywords are categorized by concept and by the amount of data available for use in predicting future behavior. The keywords and/or the categories can also be categorized as seasonal or non-seasonal. A category level seasonal variation pattern can then be calculated based on keywords in the category that have sufficient historical data. A search volume can then be predicted for one or more keywords, with an appropriate calculation algorithm being selected based on the concept category, seasonal classification, and historical data available for the keywords.