Conversion Rate Estimation Model Using Merged Attribution Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current conversion rate estimation models face challenges in accurately predicting conversion events due to insufficient training data and high resource overhead, especially during the 'cold start' phase when recommended content items are not distributed, leading to inaccurate billing and content recommendation strategies.
Innovation Solution
A method and apparatus for conversion evaluation that extracts resource features and audience features from related data, using a conversion rate estimation model trained on both attribution and non-attribution data to predict conversion probabilities, improving accuracy and distribution effectiveness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional conversion rate estimation models are used that rely only on attribution data, then the model complexity is low, but the measurement precision and reliability of conversion rate prediction deteriorate due to insufficient training data
Solution Approach 1:
The patent merges attribution data and non-attribution data into a unified training dataset. The conversion rate estimation model simultaneously processes both data types, combining their respective features (attribution features from direct conversions and non-attribution features from indirect conversions) to improve prediction accuracy while maintaining a single integrated model structure.
Solution Approach 2:
The conversion rate estimation model is designed to handle multiple data sources and multiple feature types universally. It can process both attribution and non-attribution data through the same architectural framework, making the model multi-functional and adaptable to different data conditions without requiring separate specialized models.
2Quantity of substance
If conversion rate estimation models are trained only on attribution data, then the training data quantity is limited, but the productivity and resource overhead increase due to the need for more extensive data collection and processing
Solution Approach 1:
The system performs preliminary classification of conversion data into attribution and non-attribution categories during data collection. Non-attribution conversions are identified and tagged in advance based on predefined criteria (such as conversions occurring after ad exposure but without direct attribution), preparing them for model training before the actual conversion rate prediction task begins.
3Adaptability or versatility
If the model uses only attribution data for training, then the ease of operation is high, but the adaptability to different distribution scenarios deteriorates
Solution Approach 1:
The patent segments conversion data into distinct attribution and non-attribution components, allowing the model to learn from different data characteristics separately before integrating them. This segmentation enables the model to adapt to various distribution scenarios by selectively weighting or emphasizing different data segments based on the specific context.
Data Source
AI summary
According to the embodiments of the present disclosure, a method for conversion evaluation comprises: extracting a resource feature from resource-related data of a target resource; extracting an audience feature of the target audience group from audience-related data of a target audience group of the target resource, the target audience group being to be distributed with a recommended content item related to the target resource; and determining, based on the resource feature and the audience feature, a target predicted conversion rate for the target resource through a predetermined association between resource features, audience features and predicted conversion rates, the target predicted conversion rate indicating a predicted probability of the target audience group performing a conversion for the target resource. According to the scheme, the accuracy of the conversion rate evaluation may be improved, thereby improving distribution effect of the recommended content item for a resource.