ML Recurring Transaction Prediction with Minimal History
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to accurately predict and create recurring electronic transactions, especially in B2B transactions, due to inconsistencies in periodicity and lack of information, leading to confusion for payers and payees, and existing techniques require a long history of transactions for accurate predictions.
Innovation Solution
A method and system using a machine learning model to generate features from a sequence of recurring transactions, predicting whether a subsequent transaction will occur and within what time window, allowing for the automatic creation of future recurring transactions based on these predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional automated analysis techniques are used to determine recurring transactions, then the system can identify transaction patterns, but the techniques require a long history of transactions to have accurate predictions
Solution Approach 1:
The patent changes the parameters used for analysis by extracting features related to transaction periodicity, amount consistency, and timing patterns. These transformed parameters enable accurate predictions from shorter transaction histories by focusing on the temporal and quantitative characteristics rather than requiring extensive historical data
Solution Approach 2:
The patent extracts key features from transaction data such as recurrence rate, transaction amount consistency, and timing variations. By isolating and analyzing these specific features, the system can make accurate predictions without needing to process entire long transaction histories
2Reliability
If existing techniques such as time series analysis are used, then the system can predict whether a sequence of recurring transactions will continue, but the techniques cannot simultaneously predict the next occurrence time window
Solution Approach 1:
The patent merges the classification task (predicting whether transactions will continue) with the regression task (predicting the time window for next occurrence) into a single unified model. This combined approach simultaneously provides both the continuation probability and the predicted time window, eliminating the need to choose between the two prediction types
3Device complexity
If electronic transactions do not include enough information, then the transaction data remains simple, but the system cannot determine whether the transaction is part of a sequence of recurring transactions
Solution Approach 1:
The patent segments the transaction data into distinct feature categories including temporal features (timing, periodicity), quantitative features (transaction amount), and relational features (payer-payee relationships). This segmentation allows the system to detect recurring patterns by analyzing each feature dimension independently while maintaining relatively simple overall data structure
Data Source
AI summary
The present disclosure provides techniques for recommending vendors using machine learning models. One example method includes generating a set of features based on a sequence of recurring transactions associated with a user, a payee, and a transaction amount, predicting simultaneously, using a machine learning model based on the set of features, that the sequence of recurring transactions will continue with a subsequent transaction and a time window within which the subsequent transaction of the sequence will occur, receiving electronic transaction data indicative of a transaction associated with the user, the payee, and the transaction amount, indicating that the transaction is the subsequent transaction in the sequence of recurring transactions based on the transaction and the prediction, and automatically creating one or more future recurring transactions based on the indication.