Machine Learning Model Training for Mobile Payment Risk Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In machine learning model training for mobile payment risk control, existing methods fail to accurately classify and filter out suspected positive samples within negative samples, leading to reduced model accuracy due to unreported risk events.
Innovation Solution
Classify samples with risk labels as positive samples and those without as negative samples, assess risk scores, filter negative samples with scores above a threshold, and re-train the model using the filtered samples to improve classification accuracy and model performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all negative samples are used in training without filtering, then the training data volume is maximized, but the model accuracy deteriorates due to inclusion of suspected positive samples misclassified as negative
Solution Approach 1:
The patent extracts suspected positive samples from the negative sample set by applying a risk score threshold. Negative samples with risk scores exceeding the threshold are identified and removed from the training set, thereby purifying the negative sample collection and improving model accuracy while maintaining an adequate training data volume.
Solution Approach 2:
The patent introduces risk scores as an intermediary metric to evaluate and filter negative samples. By calculating risk scores for each negative sample and comparing them against a threshold, the system mediates between maintaining training data volume and ensuring model accuracy, selectively removing only those samples that pose a risk of being misclassified positives.
2Measurement precision
If suspected positive samples are filtered out from negative samples, then the classification accuracy is improved, but the training process complexity increases due to re-training requirements
Solution Approach 1:
The patent performs preliminary filtering of suspected positive samples from negative samples before the main model training process. By pre-identifying and removing samples with high risk scores, the system prepares a cleaner training dataset in advance, which simplifies the subsequent training process and reduces the need for complex iterative adjustments during training.
Solution Approach 2:
The patent implements a feedback mechanism where the model evaluates negative samples by assigning risk scores, and these scores feed back into the sample selection process. Samples exceeding the threshold are identified and removed, creating a feedback loop that continuously improves classification accuracy while managing training complexity through systematic re-training on filtered data.
Data Source
AI summary
A machine learning model training method includes: classifying samples having risk labels in a training sample set as positive samples and classifying samples without risk labels in the training sample set as negative samples; training a risk model with a machine learning method based on the positive samples and the negative samples; obtaining a risk score for each of the negative samples based on the trained risk model; identifying one or more negative samples in the training sample set that have a risk score greater than a preset threshold value; re-classifying the one or more negative samples in the training sample set that have a risk score greater than the preset threshold value as re-classified positive samples to generate an updated training sample set from the training sample set; and re-training the risk model with the machine learning method based on the updated training sample set.