保留正文与补采附件。源页面当前未重新核验；具体限制见交付清单。

Revisit consent button We value your privacy We use cookies to enhance your browsing experience, serve personalised ads or content, and analyse our traffic. By clicking "Accept All", you consent to our use of cookies. Customise Reject All Accept All Customise Consent Preferences We use cookies to help you navigate efficiently and perform certain functions. You will find detailed information about all cookies under each consent category below. The cookies that are categorised as "Necessary" are stored on your browser as they are essential for enabling the basic functionalities of the site. ... Show more NecessaryAlways Active Necessary cookies are required to enable the basic features of this site, such as providing secure log-in or adjusting your consent preferences. These cookies do not store any personally identifiable data.

Functional Functional cookies help perform certain functionalities like sharing the content of the website on social media platforms, collecting feedback, and other third-party features.

Analytics Analytical cookies are used to understand how visitors interact with the website. These cookies help provide information on metrics such as the number of visitors, bounce rate, traffic source, etc.

Performance Performance cookies are used to understand and analyse the key performance indexes of the website which helps in delivering a better user experience for the visitors.

Advertisement Advertisement cookies are used to provide visitors with customised advertisements based on the pages you visited previously and to analyse the effectiveness of the ad campaigns.

Uncategorised Other uncategorised cookies are those that are being analysed and have not been classified into a category as yet. No cookies to display. Reject All Save My Preferences Accept All Toggle SearchToggle Mobile Navigation Publish AI, ML & data-science insights to a global community of data professionals. Sign inSubmit an Article LatestEditor's PicksNewsletterShipAI

Write For TDS LinkedInXToggle Search Artificial Intelligence

Can Synthetic Data Boost Machine Learning Performance?

Investigating the Capability of Synthetic Data to Enhance Model Performance on Imbalanced Datasets John Adeojo July 5, 20237 min read Image by Author: Generated with Midjourney Image by Author: Generated with Midjourney

Background - Imbalanced Datasets

Imbalanced classification problems frequently occur in commercial machine learning use cases. You may encounter them in churn prediction, fraud detection, medical diagnosis, or spam detection. In all these scenarios, what we aim to detect belongs to the minority class, which can be highly underrepresented in our data. There are several approaches proposed for enhancing the performance of models on imbalanced datasets:

In this article I demonstrate how training a model on synthetic data surpasses the other approaches in enhancing the performance of the classifier.

Learn this step by step with the interactive AI Engineer, AI and Data Scientist, Computer Science and Machine Learning roadmaps.

The Dataset

The data is sourced from Kaggle, consisting of 284,807 credit card transactions, 492 (0.172%) of which are labelled as fraudulent. The data is available for both commercial and non-commercial usage under the Open Data Commons license. For interested readers, Kaggle offers more detailed information and basic descriptive statistics about the data. From this Kaggle dataset, I create two subsets: a training set and a holdout set. The training set comprises 80% of the total data, along with synthetically generated samples when exploring that approach. The holdout set constitutes 20% of the original data, excluding any synthetic samples. Image by Author: Data splitting process Image by Author: Data splitting process

The Model

I use Ludwig, an open-source, declarative framework for building deep learning models due to it's ease of implementation. Models are easily built and trained by declaring them in a yaml file and running a training job through Ludwig's python API. I have written an article previously that details Ludwig for those who are interested. For each approach, I use the same baseline model, only adjusting specific parameters as necessary. For instance, Ludwig allows for weight and sampling adjustments natively - These are simply adjusted in the yaml file. I have provided links to the model configuration yaml files for each approach for your exploration.

Generating Synthetic Data

I utilise Synthetic Data Vault (SDV), an open-source library for generating synthetic data samples. With SDV, I generate an additional 284k synthetic fraud samples, thereby achieving equal representation of both classes in the training dataset. The synthetic samples are generated with variational autoencoders adapted for tabular data (TVAE). You can find more details on the theory behind TVAEs in this paper. SDV offers diagnostic statistics, giving an indication of fit quality. You can manually explore the fit quality by comparing variable distributions in the real data versus the generated data as shown in the examples below. Image by Author: Real vs synthetic distribution for variable v1 Image by Author: Real vs synthetic distribution for variable v1 Image by Author: Real vs synthetic distribution for variable v10 Image by Author: Real vs synthetic distribution for variable v10 Image by Author: Real vs. synthetic distribution for variable amount Image by Author: Real vs. synthetic distribution for variable amount

Assessing Performance with Precision Recall Charts

We assess the performance of each model by plotting the precision versus recall curves of the models against the holdout dataset.

Precision-Recall Curve

The Precision-Recall curve, a plot of Precision (on the y-axis) against Recall (on the x-axis) for varying thresholds, is akin to the ROC curve. It serves as a robust diagnostic tool for evaluating model performance in scenarios of significant class imbalance, such as our credit card fraud detection use case, a prime example. The top-right corner of the plot represents the "ideal" point - a false positive rate of zero and a true positive rate of one. A skilled model should reach this point or come close to it, implying a larger area under the curve (AUC-PR) can suggest a superior model.

No Skill Predictor

A "no skill" predictor is a naïve model that makes predictions randomly. For imbalanced datasets, the no skill line is a horizontal line at a height equivalent to the positive class proportion. This is because if the model randomly predicts the positive class, precision would be equivalent to the positive instances proportion in the dataset.

Model Performance - Baseline

The baseline model is the deep neural network with no sample adjustments, loss function adjustments, or augmented training data. Each approach is compared to the baseline performance, which serves as a performance bench mark. Image by Author: Precision-recall curve for baseline model Image by Author: Precision-recall curve for baseline model

Model Performance - Weighted Losses Approach

Weighted loss adjusts the loss function based on the ratio of fraudulent to non-fraudulent transactions. Image by Author: Precision-recall curve for loss weighted approach Image by Author: Precision-recall curve for loss weighted approach

Model Performance - Oversampling Approach

Oversampling randomly oversamples the fraudulent transactions until there is equal representation across the classes in the training dataset. Image by Author: Precision-recall curve for oversampling approach Image by Author: Precision-recall curve for oversampling approach

Model Performance - Undersampling Approach

Undersampling randomly undersamples the non-fraudulent transactions until there is equal representation across classes in the training dataset. Image by Author: Precision-recall curve for undersampling approach Image by Author: Precision-recall curve for undersampling approach

Model Performance - Synthetic Data Approach

Leverage the TVAEs to produce 284k synthetic, fraudulent samples to gain an equal representation across classes in the training dataset. Image by Author: Precision-recall curve for synthetic data approach Image by Author: Precision-recall curve for synthetic data approach

Bootstrapping Holdout Dataset

To obtain a robust view of performance on the holdout set, I created fifty bootstrapped holdout sets from the original. Running the models associated with each approach across all sets provides a distribution of performance. We can then determine whether each approach is statistically significantly different from the baseline using the Kolmogorov-Smirnov test. Weighted : The weighted approach marginally underperformed across recall and AUC relative to the baseline. In addition to this, the variance across each performance metric appears quite high relative to the other approaches. Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Weighted Loss, KS stats - AUC 0.420 p-value 0.000, precision 0.260 p-value 0.068, Recall 0.520 p-value 0.000 Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Weighted Loss, KS stats - AUC 0.420 p-value < 0.000, precision 0.260 p-value 0.068, Recall 0.520 p-value < 0.000 Oversampling : The oversampling approach improves model recall relative to baseline, but results in a drastic deterioration of the precision. Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Oversampling, KS stats - AUC 0.160 p-value 0.549, precision 1.0 p-value 0.000, Recall 0.9 p-value 0.000 Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Oversampling, KS stats - AUC 0.160 p-value 0.549, precision 1.0 p-value < 0.000, Recall 0.9 p-value < 0.000 Undersampling : The approach performs worse than baseline across all metrics. Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Oversampling, KS stats - AUC 0.880 p-value 0.000, precision 0.6 p-value 0.000, Recall 1.0 p-value 0.000 Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Oversampling, KS stats - AUC 0.880 p-value < 0.000, precision 0.6 p-value < 0.000, Recall 1.0 p-value < 0.000 Synthetic : The synthetic method uplifts model recall, albeit at the cost of precision. While the impact on precision remains substantial, the synthetic approach provides a more resilient alternative for enhancing model recall with less of a detriment to precision when compared to the oversampling approach. The robustness of the synthetic approach is further evidenced by the uplift in AUC-PR. Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Synthetic, KS stats - AUC 0.620, Precision 0.560, Recall 0.360 all p-values ≤ 0.003 Image by Author: Model performance metrics over 50 bootstrapped holdout samples. Baseline vs Synthetic, KS stats - AUC 0.620, Precision 0.560, Recall 0.360 all p-values ≤ 0.003

Conclusion

We've noted that the synthetic data approach can boost model recall relative to the baseline at the expense of precision. Oversampling accomplishes a similar result, but model precision suffers drastically in comparison. In our specific context of credit card fraud detection, false positives are not as costly as false negatives. Therefore, we can afford to compromise on model precision if it results in a significant boost in recall. Enriching our training data with synthetic instances seems to be an effective strategy to enhance recall while mitigating the detrimental effects on precision. This enhancement could notably affect profitability, especially when scaling the model to handle millions of transactions. Ultimately, attributing a exact cost to false positives and negatives will provide us with a clearer understanding of the most commercially viable approach, a topic beyond the scope of this article. It would be fascinating to examine the performance across varying sample sizes of synthetic data, perhaps in conjunction with weighted losses. Similarly, experimenting with diverse oversampling ratios could potentially yield comparable effects to what we have observed with the synthetic approach. The notebook for this project is available in my GitHub repo

Follow me on LinkedIn Subscribe to medium to get more insights from me:

Join Medium with my referral link - John Adeojo Should you be interested in integrating AI or data science into your business operations, we invite you to schedule a complimentary initial consultation with us: Book Online | Data-Centric Solutions

Register Now to Continue Reading

Unlock Towards Data Science with a Free Account

All articles on Towards Data Science are available to registered users. Create a free account today to access this article and thousands more. Register for freeSign in to your account

Written By

John Adeojo

See all from John Adeojo Editor's PicksArtificial IntelligenceData ScienceMachine LearningTechnology Share This Article

Towards Data Science is a community publication. Submit your insights to reach our global audience and earn through the TDS Author Payment Program. Write for TDS

Related Articles

Implementing ML Systems tutorial: Server-side or Client-side models?

Artificial Intelligence Coding a small Flask server vs Coding a WebDNN browser model Mostafa Ibrahim June 12, 20217 min read

Preventing Outcome Starvation

Machine Learning Why degenerate feedback loops need to be urgently addressed when operating machine learning models Anindya Datta March 21, 20229 min read

Measuring Data Ecosystem Maturity

Data Science Introduction to a scale reflecting economic value of data. Adam Votava December 7, 20215 min read

Top 3 Books to Kickstart Your Machine Learning Journey

Artificial Intelligence Wheather you are a data analyst or a programmer Dario Radečić October 16, 20204 min read

Contributing to open source for beginners

Data Science Or how I went from barely knowing how to use GitHub to 4 accepted pull requests in just a few weeks Maria Ilie November 3, 20204 min read

Arabic Topic classification On The Hespress News Dataset

Artificial Intelligence How to classify Arabic Text the right way Tariq Massaoudi October 18, 20206 min read

Modeling and Navigating a Transportation Network with Memgraph

Introduction Barbara Prkacin April 12, 202111 min read

Why is the MAD always smaller than the STD?

Programming Intuition, a derivation and Jensen's Inequality The Mean Absolute Deviation (MAD) and Standard Deviation (STD)... Mark Jamison February 17, 20227 min read

Why and how custom exceptions lead to cleaner, better code

Programming Clean up your code by creating your own custom exceptions Mike Huls November 15, 20222 min read

Your home for data science and AI. The world's leading publication for data science, data analytics, data engineering, machine learning, and artificial intelligence professionals. Subscribe to Our NewsletterWrite For TDS•About•Advertise•Privacy Policy•Terms of Use © Insight Media Group, LLC 2026•The New Stack•roadmap.sh•Creator Network

Recommended Articles

Close