Ensemble Model for Account Prediction in Support Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing support systems struggle to accurately and automatically determine which account a user is seeking support for, especially when users have multiple accounts with an application.
Innovation Solution
A machine learning-based approach using an ensemble model comprising multiple tree-based models and a ranking model is employed to predict the account related to a support request. The model is trained on historical support records with features derived from time-stamped actions and support initiation times.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional support systems are used to determine account identity, then the system structure is simple, but the accuracy of automatic account determination deteriorates when users have multiple accounts
Solution Approach 1:
The support system is segmented into multiple specialized components: an ensemble model containing multiple tree-based models for different analysis tasks, a ranking model for prioritizing accounts, and feature engineering modules for processing different data types. Each component handles specific aspects of account determination, improving overall accuracy while maintaining manageable complexity through functional decomposition.
Solution Approach 2:
The system employs a composite modeling approach by combining multiple tree-based models (Random Forest, Gradient Boosting, XGBoost) with a ranking model in an ensemble architecture. This composite structure leverages the strengths of different algorithms to achieve superior account determination accuracy compared to any single model, while the modular design keeps the system complexity manageable.
2Productivity
If manual account identification is used, then the system complexity is low, but the support efficiency and time to identify accounts deteriorates
Solution Approach 1:
The system implements self-service by automatically determining account identity without requiring manual intervention. The ensemble model autonomously processes support requests, analyzes user behavior patterns, and identifies the relevant account, significantly improving support efficiency. The automated nature eliminates the need for support agents to manually investigate account associations.
Solution Approach 2:
The manual mechanical process of account identification is replaced with an automated machine learning system. The ensemble model processes data and makes account determination decisions algorithmically, substituting human analytical work with computational processes that are faster and more consistent, thereby improving productivity.
3Measurement precision
If basic user identification is used, then the system is easy to operate, but the ability to determine the correct account among multiple accounts deteriorates
Solution Approach 1:
The system performs preliminary actions by pre-processing support request data and user behavior data before account determination. Feature engineering pipelines prepare data in advance, and the ensemble model is pre-trained on historical data, enabling accurate account identification without requiring complex real-time processing during support interactions, thus maintaining ease of operation.
Solution Approach 2:
The system introduces an intermediary layer of feature engineering and data processing between the raw support request and the account determination decision. This intermediary layer transforms complex raw data into meaningful features that the ensemble model can process effectively, improving account identification accuracy while shielding the operational interface from complexity.
4Measurement precision
If an ensemble model with multiple tree-based models is used, then the account prediction accuracy is improved, but the computational resources and model complexity increase
Solution Approach 1:
The system applies partial action by using multiple tree-based models selectively for different aspects of account determination rather than applying all possible analysis methods uniformly. The ensemble structure allows the system to leverage multiple models only when needed for complex multi-account scenarios, reducing unnecessary computational resource consumption while maintaining high accuracy when required.
Data Source
AI summary
Aspects of the present disclosure provide techniques for training a machine learning model. Embodiments include receiving a historical support record comprising time-stamped actions, a support initiation time, and an account indication. Embodiments include determining features of the historical support record based at least on differences between times of the time-stamped actions and the support initiation time. Embodiments include determining a label for the features based on the account indication. Embodiments include training an ensemble model, using training data comprising the features and the label, to determine an indication of an account in response to input features, wherein the ensemble model comprises a plurality of tree-based models and a ranking model.