Bond Price Prediction via Dynamic Clustering and Error Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for predicting bond prices using AI models, such as Hidden Markov Models (HMM), face challenges in achieving accurate predictions and expanding coverage to include a wider range of bonds, due to issues like outliers and low intra-cluster correlation.
Innovation Solution
The proposed solution involves constructing structured datasets by reclustering bonds based on error analysis and correlation coefficients, using a combination of supervised and unsupervised machine learning techniques to improve prediction accuracy and expand bond coverage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional clustering methods are used to group bonds for AI model training, then the model can process bonds in groups, but the intra-cluster correlation remains low and prediction accuracy is limited
Solution Approach 1:
The patent segments bonds into clusters based on price movement similarity rather than traditional classification. By dividing the bond dataset into groups with high intra-cluster correlation (bonds that move together in price), the system improves prediction accuracy while maintaining reliable cluster structures. This segmentation approach directly addresses the contradiction by creating meaningful groups that enhance both measurement precision and reliability.
Solution Approach 2:
The patent implements dynamic reclustering that adapts to changing market conditions. The clustering is not static but dynamically adjusted based on recent price movements and error analysis. This dynamic approach allows the system to maintain high intra-cluster correlation even as market conditions evolve, thereby improving prediction accuracy while preserving cluster reliability over time.
2Adaptability or versatility
If AI models are trained on existing bond datasets, then price predictions can be generated, but coverage is limited and error rates remain high for certain bond types
Solution Approach 1:
The patent implements a feedback mechanism where prediction errors are analyzed and used to improve future predictions. By identifying bonds with high error rates and analyzing why predictions failed, the system adjusts its clustering and modeling approaches. This feedback loop expands bond coverage by learning from mistakes while simultaneously reducing error rates through continuous improvement of the prediction model.
Solution Approach 2:
The patent changes key parameters such as clustering criteria, time window sizes, and model hyperparameters based on performance analysis. By dynamically adjusting these parameters, the system expands its ability to handle diverse bond types (increasing coverage) while optimizing for accuracy (reducing error rates). Parameter changes allow the model to adapt to different bond characteristics and market conditions.
3Adaptability or versatility
If more bonds are included in the prediction model to expand coverage, then versatility improves, but prediction accuracy and reliability may deteriorate due to outliers and heterogeneity
Solution Approach 1:
The patent uses segmentation to divide heterogeneous bond data into homogeneous clusters. By grouping bonds with similar price movement patterns together, the system can include a wide variety of bond types (expanding coverage) while maintaining reliable predictions within each cluster. The segmentation ensures that outliers don't negatively impact the entire model but are handled appropriately within their own clusters.
Data Source
AI summary
Aspects of the subject disclosure may include, for example, systems and methods for generating structured datasets for predicting bond price. The systems and methods include constructing a price function of each bond contained in a plurality of bond clusters including a target cluster, training a machine learning model to determine a cause for an erroneous price prediction result, and generating structured datasets based on a feedback from the machine learning model. Other embodiments are disclosed.