Quantum Computing in Finance: The Possible Use Cases

We always talk about use cases at Quantum Zeitgeist. Some of those early use cases are likely to be in finance and pharmaceuticals. Here we take a deeper look at some of the techniques that developers and researchers are looking at in the financial space. Of course, Quantum Machine Learning is one of the “hottest” topics in Quantum Computing, which just like classical machine learning, is unlocking the promise of automation.

Regression

Regression is concerned with training a model to estimate real-valued functions. This has a very useful application in experimental sciences and forecasts. By using quantum computers for regression, Parameterized Quantum Circuits can be trained as function estimators by finding the mathematical relationship between two variables and using the value of one of the variables to predict the measurement of the other variable.  

One of the common applications of classical regression technique is in learning time series through Recurrent Neural Networks. This technique is commonly used to make predictions about evolving processes from historical data. There has also been a proposed improvement in convergence and error rate through the use of PQCs as function models in Quantum versions of Recurrent Neural Networks (RNNs).

Financial applications of regression techniques using quantum computers include:

  • Asset Pricing: This has to do with assigning prices to different categories of financial instruments such as stocks, bonds, etc. Historical financial data through regression techniques can be used to evaluate and more accurately predict spot prices.
  • Multi-Asset Trend Following Strategy: Classical Regression models can be used to predict 1-day returns of a multi-asset class portfolio. These returns can be predicted by using historical prices at different times to create a trend that can be used to predict future prices.
  • Implied Volatility Estimation: Regression Technique is used to analyse the possibility of changes in the price of a given security in the financial market. The proposed quantum approach for this application uses the deep quantum neural network.

Classification

Classification is concerned with predicting labels for new data points using a model that is fit by a labelled dataset. Some classical methods are used for classification algorithms: Linear Classification, Nearest Centroid, Support Vector Machines (SVM). Based on recent research, Neural Network-based methods have been proven to have better performance than traditional methods. When a Neural Network’s weight is trained using a labelled dataset, it can be used to perform inference on unseen data instances.

Discussing the techniques under Classification; a Linear Classifier is used to classify objects in a dataset based on the value of a linear combination of the characteristics of the objects referred to as feature values and stored as a feature vector.

Distance-Based Classifiers such as Nearest Centroid and K-nearest neighbours are used to predict the label of a new data point based on its distance to reference points. Quantum approaches of the Nearest Centroid and K-NN have been proposed to enhance performance. The quantum version of the Nearest Centroid Algorithm was used to perform Classification in the Modified National Institute of Standards and Technology (MNIST).

Support Vector Machines (SVM) relates to solving a convex quadratic optimization problem to find the hyperplane that results in the maximum margin between two classes of data.

Additionally, Variational Quantum Classifier is another technique used under Classification

Financial applications of classification techniques using quantum computers include:

  • Binary Options Reduction: The Support Vector Machine technique can be used to predict the outcome of exotic options. This is the type of exotic option often used in foreign exchange. They are used to separate two classes corresponding to a binary option outcome.
  • Financial Forecasting: This tool is used to adjust to uncertainty based on predictions. Techniques such as K-NN have proposed models that match a company’s recent earning to the historical earnings of competitors to predict future earnings of the company.
  • Credit Scoring: Algorithms using weighted K-NN has been proposed for credit scoring. This is very useful for banks and loan companies. It is used to predict the credit risk of an application.

Clustering

This involves identifying groups of data points that are close to each other according to certain metrics. Quantum Clustering belongs to the family of density-based algorithms where clusters are defined by regions of higher density of data points. 

Dynamic quantum clustering, an improved version of quantum clustering adopts a time-dependent Schrodinger equation to study the evolution of quantum states associated with data points and the structure of the potential energy function.

Financial applications of Clustering techniques using quantum computers include:

  • Fraud Detection: Clustering can be used to detect anomalies through improved learning from the imbalance dataset. This feature is used to detect anomalies in new data points.
  • Stock Selection: Clustering helps investors to track stocks with similar returns but different risks. Therefore, helping investors to maximize profit and minimize risk by using the same return stock but with minimal risk.
  • Exchange Rate Regimes: K-mean technique can be used to perform cluster analysis for exchange rate regimes.
  • Hedge Fund Clustering: Clustering methods such as K-means have been used to overcome the challenges faced in Hedge Fund Clustering.

Generative Modelling

This model learns a probability distribution over data. Since measuring a quantum state naturally results in a probability distribution over the outcomes, it makes sense to see if quantum computation can be utilized for generative modelling. 

Some of the techniques of generative modelling include Boltzmann Machine, Generative Adversarial Learning, and Quantum Born Machine.

Financial applications of Generative Modelling techniques using quantum computers include:

  • Fraud Detection: Quantum versions of the Boltzmann Machine have been especially used to detect fraud by classifying anomalous credit card transactions. They have also been used for other generative learning and discriminative learning task.
  • Probability Distribution Preparation: Quantum machines have been used to learn PQCs for loading probability distributions. The use of these quantum machines helps in the efficient preparation of input probability distribution.

Quantum Assisted Feature Extraction

The set of techniques that are useful in identifying the attributes of a dataset such as regression and classification are known as Feature extraction. Quantum algorithms enhance this task by helping to compute properties of the dataset that a classical computer will ordinarily find difficult to identify or take a long time to do so. Feature extraction is used in detecting anomalies in transactions. Tools such as graph-theoretic tools are being used to study bidding markets and identify colluding communities or cartels.

Financial applications of Quantum Assisted Feature Extraction techniques using quantum computers include:

  • Model Reduction: The Probability Component Analysis technique (PCA) can be used to ease model tuning. This is done when the model needs to be tuned to match the implied volatility, volatility estimated by the model with the actual market volatility. IT helps to reduce the number of components and parameters which in turn eases model tuning.
  • Combinatorial Feature Selection for Credit Score Classification

Reinforcement Learning

This is a machine learning technique where an agent attempts to learn through interactions with the environment. According to the approach utilized by Dong et al in 2005, the possible actions at any given state in the environment are maintained in a quantum superposition, and amplitude amplification is used to increase the probability of measuring a good action at any given state.

Financial applications of Reinforcement Learning techniques using quantum computers include:

  • Algorithmic Trading: This has to do with the automatic execution of trades of a financial instrument without human intervention. This is performed by predictions in a supervised manner followed by obtaining optimal trading decisions under uncertainty associated with the corresponding predictions and market volatility.
  • Market Making: Reinforcement Learning is essential for market makers who have the role of maintaining a set of sell and buy orders at various quantities and prices. They make a profit from the market through the market gap called the spread.

Natural Language Processing

This is concerned with automated text and language analysis. Classical methods of Natural Language Processing encounters challenges in understanding separate words alone and not grammatical structures. The improved version of this technique uses variational PQCs to encode classical data on quantum hardware can encode linguistic structures faster than its classical counterpart.

Financial applications of Natural Language Processing techniques using quantum computers include:

  • Risk Assessment: Natural Language Processing techniques can be used to quantify the chances of a successful loan payment based on credit risk assessment. This technique can measure attitude and entrepreneurial mindset in business loans.
  • Financial Forecasting: Natural Language Processing techniques are frequently used for financial forecasting. It also helps in carrying out sentiment analysis which is an important role in decision making by traders.
  • Accounting and Auditing: Natural Language Processing is also applied in accounting and auditing with the major aim of detecting fraud through the evaluation of accounting systems, assessment of risk fraud, monitoring of internal control, etc.

Seven techniques were explained and their applications in the financial sector were highlighted. The techniques outlined in more detail demonstrated that quantum computers have the potential of giving more accurate performances than classical computers when used for numerous financial applications.