<p>ML usage in wealth management is a growing trend. Approximately 50% of S&amp;P 500 earnings calls have included references to AI and ML since May 2023. Asset and wealth management is yet in the early adoption stage of ML.<br />
In this blog, we’ll review some of the most beneficial techniques for the A&amp;WM industry.</p>
<p><b>Key AI/ML techniques</b></p>
<table>
<tbody>
<tr>
<th style="border: 1px solid black; padding: 10px;"><b>Technique</b></th>
<th style="border: 1px solid black; padding: 10px;"><b>Characteristics</b></th>
<th style="border: 1px solid black; padding: 10px;"><b>Typical application</b></th>
</tr>
<tr>
<td style="border: 1px solid black; padding: 10px;">Artificial neural networks</td>
<td style="border: 1px solid black; padding: 10px;">Nonlinear regression model<br />
Network of connected nodes that loosely model neurons in a brain<br />
Receives a training set of input and desired output data pairs and can learn the relationship between them<br />
Can then be used to predict the output of previously unseen inputs</td>
<td style="border: 1px solid black; padding: 10px;">Forecasting</td>
</tr>
<tr>
<td style="border: 1px solid black; padding: 10px;">Decision trees and random forests</td>
<td style="border: 1px solid black; padding: 10px;">A decision tree classifies units based on their features<br />
Classification is done by traversing a logical tree from root to leaves, with each branch moving left or right depending on the unit’s features; such trees can be interpreted by humans<br />
Constructed automatically based on the training set input and desired output pairs<br />
Random forests simply average the outputs of several decision tree models to produce more reliable forecasts</td>
<td style="border: 1px solid black; padding: 10px;">Classification and forecasting</td>
</tr>
<tr>
<td style="border: 1px solid black; padding: 10px;">Support vector machines</td>
<td style="border: 1px solid black; padding: 10px;">Can be used for classification or regression<br />
Can handle nonlinear relationships by mapping the inputs to a higher-dimensional space<br />
Faster to train than artificial neural networks</td>
<td style="border: 1px solid black; padding: 10px;">Forecasting</td>
</tr>
<tr>
<td style="border: 1px solid black; padding: 10px;">LASSO</td>
<td style="border: 1px solid black; padding: 10px;">Ordinary regression model with an additional penalty tern that ensures choosing the smallest necessary subset of explanatory variables<br />
Reduces spurious coefficient estimates to zero, which significantly enhances the out-of-sample performance of the model</td>
<td style="border: 1px solid black; padding: 10px;">Forecasting</td>
</tr>
<tr>
<td style="border: 1px solid black; padding: 10px;">Cluster analysis</td>
<td style="border: 1px solid black; padding: 10px;">Cluster data into groups so that the units in each group have similar characteristics<br />
The number of clusters can be defined by the user or determined automatically by the algorithm</td>
<td style="border: 1px solid black; padding: 10px;">Asset classification</td>
</tr>
<tr>
<td style="border: 1px solid black; padding: 10px;">Evolutionary (Genetic) algorithms</td>
<td style="border: 1px solid black; padding: 10px;">Optimization technique capable of searching through large, complex, nonlinear sets of solutions, identifying those that preferred<br />
Process inspired by natural evolution</td>
<td style="border: 1px solid black; padding: 10px;">Variants of portfolio optimization that can not be solved with classical optimization algorithms</td>
</tr>
<tr>
<td style="border: 1px solid black; padding: 10px;">Natural language processing</td>
<td style="border: 1px solid black; padding: 10px;">Range of techniques used to process natural language data (textual, audio)<br />
Particularly useful for extracting information from textual media, like social media, websites, news articles</td>
<td style="border: 1px solid black; padding: 10px;">Automatic analysis of corporate annual reports and news articles</td>
</tr>
</tbody>
</table>
<h2>AI and ML in asset management</h2>
<p>AI can be used to construct portfolios that meet performance targets better than those created using traditional methods.</p>
<div id="attachment_137989" style="width: 1034px" class="wp-caption alignnone"><img loading="lazy" decoding="async" aria-describedby="caption-attachment-137989" class="wp-image-137989 size-large" src="https://dev.sombrainc.com/app/uploads/2024/01/ai-portfolio-management-1024x637.png" alt="AI in portfolio management" width="1024" height="637" /><p id="caption-attachment-137989" class="wp-caption-text">AI in portfolio management</p></div>
<h3>Portfolio management</h3>
<h4>NLP approaches</h4>
<p><b>NLP approaches</b> are capable of extracting economically meaningful information from various sources of text, such as corporate annual reports, news articles, and Twitter posts. AI approaches can also interpret context and sentence structure.<br />
<b>Traditional approach</b> is basically manual labor. An advisor has to analyze dozens, if not hundreds of 20-30-paged reports.<br />
NLP can also be used for:</p>
<ul>
<li>asset revenue and risk estimation,</li>
<li>portfolio creation from scratch,</li>
<li>rebalancing using derivatives.</li>
</ul>
<p>Using AI, the advisor is able to boost their performance and give optimal advice faster.</p>
<div id="attachment_137990" style="width: 1034px" class="wp-caption alignnone"><img loading="lazy" decoding="async" aria-describedby="caption-attachment-137990" class="wp-image-137990 size-large" src="https://dev.sombrainc.com/app/uploads/2024/01/sample-machine-learning-1024x637.png" alt="Example of how to analyze data sets to identify signals" width="1024" height="637" /><p id="caption-attachment-137990" class="wp-caption-text">Example of how to analyze data sets to identify signals</p></div>
<h5>Morgan Stanley use case</h5>
<p>For example, earlier this year, Morgan Stanley deployed OpenAI’s GPT-4 to empower their wealth management advisors with a new way to utilize the company’s vast intellectual capital. Advisors can get answers to specific questions by asking the AI instead of searching through PDFs on internal websites. The AI provides plain English answers with links to source documents.</p>
<blockquote><p>“You essentially have the knowledge of the most knowledgeable person in Wealth Management—instantly,” says Jeff McMillan, the company’s Head of Analytics, Data &amp; Innovation</p></blockquote>
<h5>Sombra case study</h5>
<p>After more than a decade of a low-interest rate environment, inflation picked up, and the Fed responded with an increase in interest rates. However, since September 2023, the market data remains mixed, with no clear sight of the following FOMC steps.</p>
<p>Major Bulge Bracket Banks are divided on their view on the future rate path. It’s puzzling to see traders betting the seventh time’s the charm when it comes to the Fed pivoting from rate hikes to cuts, given that officials have repeatedly said their fight against inflation is far from over.</p>
<p><b>Business challenge.</b> Since March 2022, almost all the client’s asset classes have performed poorly while the correlation between fixed-income assets and equities has surged, rendering fixed-income ineffective in its role as a hedging tool. As a result, a classic 60/40 portfolio resulted in roughly the same performance as the S&amp;P 500 Index.</p>
<p><b>Sombra approach.</b> We applied Loughran-McDonald sentiment word lists and BERT and XLNet ML techniques for NLP to FOMC statements to see if they anticipated changes in the federal funds rate and then examined whether our results had any correlation with stock market performance.</p>
<div id="attachment_137991" style="width: 1034px" class="wp-caption alignnone"><img loading="lazy" decoding="async" aria-describedby="caption-attachment-137991" class="wp-image-137991 size-large" src="https://dev.sombrainc.com/app/uploads/2024/01/nlp-flow-sombra-1024x637.png" alt="Sombra NLP flow" width="1024" height="637" /><p id="caption-attachment-137991" class="wp-caption-text">Sombra NLP flow</p></div>
<p><b>Value for the client.</b> With the value of asset diversification diminished and global investment firms polarized on their view on the further Fed rate movement, achieving an objective and quantifiable understanding of the Federal Open Market Committee’s (FOMC) outlook has grown ever more critical.</p>
<p>Start capitalizing on machine learning opportunities today with Sombra’s wealth management services. Let us know what your business problem is</p>
<h4>The LASSO framework</h4>
<p><b>The LASSO framework</b> can also be used to find lead–lag relationships between asset groups or markets.<br />
<b>Traditional approach</b> here embodies a regular linear regression, which shows the correlation between the assets based on historical data. The main disadvantage of this approach is regression results being sensitive to:</p>
<ul>
<li>Timespan of the historical data. The regression of the last 10 and 20 years will have different values.</li>
<li>Date of the analysis’ start. The correlation value would be higher at the beginning of a crisis and lower during a calm environment.</li>
</ul>
<h4>Neural networks</h4>
<p><b>Neural networks</b> are one of the most popular AI techniques for predicting stock returns, company fundamentals, and returns of other asset classes such as bonds.<br />
<b>Traditional approach</b> lies in manual historical data observation. This approach shares the same drawbacks as using linear regression instead of the LASSO framework, apart from</p>
<ul>
<li>manually building financial models (DCF),</li>
<li>being subjective,</li>
<li>time-consuming.</li>
</ul>
<h5>Nvidia case study</h5>
<p>A WM research firm’s technology stack generates return-predictive signals for thousands of securities, multiple times a day. Powered by multiple DGX Stations and GPUs. Transitioning from CPU to GPU HPC environment significantly accelerated R&amp;D &#8211; <b>cutting computation time by more than 80%.</b></p>
<div id="attachment_137992" style="width: 1034px" class="wp-caption alignnone"><img loading="lazy" decoding="async" aria-describedby="caption-attachment-137992" class="wp-image-137992 size-large" src="https://dev.sombrainc.com/app/uploads/2024/01/stock-prediction-flow-1024x637.png" alt="Stock prediction flow" width="1024" height="637" /><p id="caption-attachment-137992" class="wp-caption-text">Stock prediction flow</p></div>
<h4>Artificial neural networks</h4>
<p><b>ANNs</b> can be used for pricing and hedging using nonparametric option pricing frameworks that perform better than the Black–Scholes model in terms of delta hedging and forecasting future option prices.</p>
<p>The drawback of traditional approach includes</p>
<ul>
<li>too many variables (greeks),</li>
<li>the variables change constantly,</li>
<li>requires recalculation.</li>
</ul>
<p>ANNs, on the contrary, can also be used for rebalancing using derivatives. They ease the estimation of delta and gamma hedging.</p>
<h3>Portfolio optimization</h3>
<p>AI can generate more accurate returns and risk estimates for traditional portfolio construction.<br />
The main advantage is that AI/Ml techniques promise a higher risk-adjusted return.</p>
<p>Additionally, AI techniques can provide alternative portfolio construction approaches to generate more accurate portfolio weights and produce optimized portfolios with better out-of-sample performance than those generated.</p>
<p>Have an idea to implement ML in your asset management business? Let’s start this project together!</p>
<h3>Trading</h3>
<p>Algorithms can play a role in all stages of the trading process.</p>
<div id="attachment_137993" style="width: 1034px" class="wp-caption alignnone"><img loading="lazy" decoding="async" aria-describedby="caption-attachment-137993" class="wp-image-137993 size-large" src="https://dev.sombrainc.com/app/uploads/2024/01/algorithmic-trading-ai-1024x637.png" alt="Algorithmic trading with AI" width="1024" height="637" /><p id="caption-attachment-137993" class="wp-caption-text">Algorithmic trading with AI</p></div>
<h3>Portfolio risk management</h3>
<p>AI also has applications in risk management, with regard to both market risk and credit risk.</p>
<div id="attachment_137994" style="width: 1034px" class="wp-caption alignnone"><img loading="lazy" decoding="async" aria-describedby="caption-attachment-137994" class="wp-image-137994 size-large" src="https://dev.sombrainc.com/app/uploads/2024/01/artificial-intelligence-risk-management-1024x637.png" alt="AI application in risk management" width="1024" height="637" /><p id="caption-attachment-137994" class="wp-caption-text">AI application in risk management</p></div>
<h5>Gamestop case study</h5>
<p>Many funds began monitoring social media and tracking sentiment of specific stocks after the Gamestop mania hit Wall Street.</p>
<div id="attachment_137996" style="width: 1034px" class="wp-caption alignnone"><img loading="lazy" decoding="async" aria-describedby="caption-attachment-137996" class="wp-image-137996 size-large" src="https://dev.sombrainc.com/app/uploads/2024/01/Gamestop-timeline-1024x637.png" alt="Gamestop timeline of events" width="1024" height="637" /><p id="caption-attachment-137996" class="wp-caption-text">Gamestop timeline of events</p></div>
<p>Primarily, the funds used NLP techniques to do social listening. Sombra engineers can help you utilize financial opinions, posts, and reports in the most efficient way using machine learning techniques.</p>
<h2>Conclusion</h2>
<p>The A&amp;WM industry would only benefit from the adoption of AI/Ml techniques. Most significantly from:</p>
<ul>
<li>NLP approaches</li>
<li>The LASSO framework</li>
<li>Neural networks</li>
<li>ANNs</li>
</ul>
<p>The application varies from portfolio creation and optimization to risk management and trading.<br />
At Sombra, we offer ML and AI services to asset and wealth management companies seeking to boost the efficiency of their software products. To learn more &#8211; contact us.</p>


<p>60% of our clients have been working with us for more than three years. They are some of the most recognized players in logistics, finance, healthcare and e-commerce, like Wineshipping, OANDA,</p>
<p>One of the top 10 banks in the world has been our client for a long time. We have been working together for 7 years and have consistently demonstrated our expertise, which has led to an increase in the number of engineers working on the client’s projects from 1 to over 60.</p>
<p>You can also check our Clutch page for real reviews.</p>


How do I know Sombra is a reliable software development vendor?

<p>Apart from Wealth management software solution assessment, dedicated team, product discovery, and project-based development, Sombra’s full list of services includes:<br />
Software consulting and modernization services<br />
Software design and planning services<br />
Software development services<br />
Data engineering and analytics<br />
DevOps, testing and support services<br />
Staff &amp; team augmentation services<br />
For detailed information, contact us</p>


What other services does Sombra provide?

<p>At Sombra, we are committed to ensuring safe access and use of data in compliance with data protection standards and regulations.<br />
We comply with the following security standards:<br />
ISO 27001<br />
ISO 9001:2015<br />
PCI DSS<br />
GDPR<br />
SOC 2</p>


AI and ML in Asset Management: 3 Use Cases You Need to Know

FAQ

How do I know Sombra is a reliable software development vendor?

What other services does Sombra provide?

Do you have an established security policy?

Contact Us