Machine Learning Acquisition Prediction Using Bipartite Knowledge Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting business acquisitions and growth rates rely heavily on human judgment, which is prone to errors due to the complexity and volume of data involved, limiting the ability to consider multiple business entities and dimensions effectively.
Innovation Solution
A computer-implemented method using machine learning techniques to analyze financial and news data from multiple sources, generating feature vectors, creating dynamic and static bipartite knowledge graphs, and applying supervised machine learning models to predict future business acquisitions by encoding graph embedding vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If human experts perform acquisition predictions based on experience and prior knowledge, then the predictions can incorporate business intuition and contextual understanding, but the process is prone to errors due to data complexity and volume, and limited in the number of business entities and dimensions that can be considered
Solution Approach 1:
The patent replaces the mechanical human decision-making process with an automated machine learning system. The system uses supervised learning models to process financial data items and news items, transforming qualitative human judgment into quantitative automated predictions. This substitution eliminates human error while maintaining the ability to incorporate contextual understanding through carefully selected features and dimensions.
Solution Approach 2:
The patent creates a virtual model that copies and simulates human expert judgment processes. By training supervised learning models on historical acquisition data and expert annotations, the system replicates human decision-making patterns at scale. The model learns to weigh various business entities and dimensions similarly to human experts but can process vastly more data without fatigue or error.
2Adaptability or versatility
If humans perform acquisition predictions, then the process can be performed with existing expertise, but it is risky to make crucial decisions on bleak guess work and limited in the scope of business entities considered
Solution Approach 1:
The patent extends the analysis from traditional single-dimension financial metrics to multiple dimensions including financial data items, news items, and various business entity attributes. The system processes data across numerous dimensions simultaneously, allowing comprehensive consideration of many business entities while maintaining prediction reliability through supervised learning that validates results against historical outcomes.
Solution Approach 2:
The patent creates a universal prediction system that can analyze any business entity across multiple dimensions simultaneously. The supervised learning model is designed to handle diverse data types and business scenarios, making the system adaptable to different industries and acquisition contexts while maintaining consistent reliability standards through standardized evaluation metrics.
3Loss of information
If more data is involved in human prediction processes, then more comprehensive analysis is possible, but errors increase due to the complexity and volume of data
Solution Approach 1:
The patent replaces human data processing with automated computational systems that can handle large volumes of financial data items and news items without error. The system systematically processes all relevant information while maintaining accuracy through algorithmic consistency and validation against ground truth labels from historical acquisition data.
Data Source
AI summary
Embodiments provide methods and systems for determining prospective acquisitions among business entities using machine learning techniques. Method performed by server system includes accessing financial data items and news items associated with finances of business entities from data sources for particular time duration. Method includes generating financial and news feature vectors corresponding to business entities and applying machine learning models over financial feature vectors and news feature vectors associated with business entities for determining candidate set of business entities predicted to be engaged in business acquisition in future. Method includes creating dynamic bipartite knowledge graph for each distinct time durations within particular time duration and generating static bipartite knowledge graph based on dynamic bipartite knowledge graphs for distinct time durations. Method includes predicting occurrence of acquisition of at least one business entity of candidate set of business entities based on supervised machine learning model and static bipartite knowledge graph.


