Here’s the converted article in Markdown format:
Quantum Computing Powers Fraud Detection: Deloitte Study Reveals Promising Results
A recent study by Deloitte has demonstrated the potential of quantum computing in fraud detection, showcasing a hybrid neural network that outperforms its classical counterpart. The research utilized Amazon Braket, a fully managed quantum development environment, to train and test a model that accurately identified fraudulent transactions.
Training the Model
The team trained a hybrid quantum neural network using 30 epochs, with a batch size of 200, on a dataset comprising 12,000 samples. The model consisted of a classic layer and a quantum layer, which were updated at each epoch to achieve higher accuracy. At the end of training, the model demonstrated a loss of 0.0353 on the training set and 0.0119 on the validation set.
Model Results and Analysis
To evaluate the model’s performance, the team made predictions on a test set using the trained model. The neural network, being a regression model, outputted a 2-D array for each record in the test set, which needed to be converted into binary classification. A threshold was applied to determine whether the output should be classified as “No fraud” or “Fraud.” The optimal threshold value was chosen based on precision.
Here are the key results:
- Almost perfect performance on the majority class (“No fraud”)
- Good performance on the minority class (“Fraud”)
- Quantum model’s superiority in identifying fraudulent transactions more accurately
In comparison, a classic neural network model demonstrated similar performance but with slightly lower accuracy.
Challenges and Opportunities
The study highlighted three key challenges associated with quantum computing:
- Computational errors
- Increased calculation errors during the sampling phase
- Multidimensional aspects of qubits demanding a nuanced approach to assessing their computational prowess
To address these challenges, researchers are developing techniques such as error mitigation and error suppression. The potential gains offered by quantum technology in fraud detection are substantial, with the ability to revolutionize critical financial processes.
Conclusion
The Deloitte study demonstrates the promising potential of quantum computing in fraud detection, highlighting the need for organizations to invest in this emerging technology. By leveraging Amazon Braket and other quantum development environments, companies can accelerate their learning journey and capitalize on the transformative power of quantum computing.
About the Authors
Federica Marini is a Manager at Deloitte Italy’s AI & Data practice, specializing in AI, GenAI, ML, and data-driven solutions. Matteo Capozi is a Data and AI expert at Deloitte Italy, with expertise in designing and implementing advanced AI and GenAI models and quantum computing solutions. Kasi Muthu is a senior partner solutions architect at AWS, focusing on generative AI and data. Kuldeep Singh is a Principal Global AI/ML leader at AWS, with over 20 years of experience in tech.
Connect with Deloitte
To pilot this solution for your enterprise on AWS, connect with Deloitte today.