AI in Finance Summit AI in Finance Summit schedule

Applying alternative data to quantitative equity strategies has high potential and unique challenges. In this talk, we will use Delphia's machine learning driven long-short equity market neutral strategy as context to discuss the following: - Case studies to highlight the advantages of alternative data in investing in general. - The promise of alternative data in quantitative equity strategies. - The challenges in working with alternative data in Delphia's strategy.

The last decade has witnessed a large-scale adoption of Artificial Intelligence and Machine Learning (AI/ML) models in finance. Although there are many benefits that AI/ML can bring to financial services (e.g., higher accuracy, automation), it could also introduce new and amplify existing risks. In this respect, financial regulators around the world are currently working on regulatory requirements that AI/ML models should satisfy when applied by financial institutions. In this presentation we discuss some most recent developments on AI/ML model risk management.

With a growing number of sensors and data collected about individuals, the resolution of risk scoring models has become increasingly crisp. While traditional data fields such as income, educational attainment, and civil status are still used, alternative data sources are on the rise. The list of emerging startups that scrape and sell alternative data sources continues to grow. Maybe the size of your screen, your phone provider, or your email domain are not very strong predictors of risk when evaluated independently, but together can sculpt a digital persona with a very specific risk profile.

What data sources are companies using to evaluate the risk of individuals or even networks of individuals? How do we feel about data from health sensors (smartwatches, etc) to be used for risk profiling? Is this a good thing or a bad thing? Where do we draw the line?

FinRegLab is currently undergoing empirical research together with Stanford University to evaluate the implications of using machine learning with the use of models that take into account requirements of U.S. law (fair lending and adverse action), and general robustness issues. The empirical research is getting to answer questions around the explainability of credit underwriting models.

Customer Lifetime Value, CLV, is a popular measure to understand the future profitability of customers

to allocate resources in more efficient ways to keep the company alive during difficult economic

situations. We use machine learning tools to predict the expected revenue from each customer during

one year of his/her relationship with the institution as the CLV of the customer. The approach is

implemented on two datasets from two international financial institutions. Different feature engineering techniques were applied to improve the prediction power of the model. We used two stage or three stage prediction models. In the second phase, we train a reinforcement learning algorithm based on the history of marketing activities and the CLV as the state of customers to determine the optimum marketing action for customers in each state to maximize their profitability.

Using Deep Learning with Word Embeddings to improve Customer Satisfaction

Understanding customer satisfaction in retail banking requires exploring and comprehending multiple sources of feedback, such as emails, social networks reviews, web feedback, bot interactions, as well as speech-to-text transcripts collected from call centers. Since such a vast amount of textual data can be difficult to leverage with traditional text mining techniques, deep learning and word embeddings can be used to automatically classify and label feedback, and then deeply analyze and understand their content. In this communication we explain how we leverage all those AI techniques to get an in-depth understanding of the opinions and needs of National Bank’s retail customers. We also show how we improve the performance levels of those AI tools using in-house algorithms and data resources to improve the overall capacity of natural language understanding.

Key Takeaways: • Industrial applications • State of the art classification • Understanding of DL embedding limits

Eric Charton holds a Master in machine learning applied to voice recognition, and a Ph.D. in machine learning applied to Information extraction and natural language generation. He worked as scientist and research project coordinator in academic context in Europe (University of Avignon) and North America (CRIM, École Polytechnique de Montréal) before becoming head of search engine research and development at Yellow Pages Canada. Since March 2018, he is Senior AI Director at National Bank of Canada.